Discovery and modulation of acid-sensing ion channel modulating venom peptides Ben Cristofori-Armstrong Bachelor of Science (Hons)

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2018

Faculty of Medicine, School of Biomedical Sciences

I Abstract

Acid-sensing ion channels (ASICs) are a family of proton-activated cation channels expressed in a variety of neuronal and non-neuronal tissues. As proton-gated channels, they have been implicated in many pathophysiological conditions where pH is perturbed. Venom derived compounds represent the most potent and selective modulators of ASICs described to date, and thus have been invaluable pharmacological tools to study ASIC structure, function, and biological roles. There are now eleven ASIC modulators described from animal venoms, with those from snakes and spiders favouring ASIC1, while the sea anemone and cone snail modulators preferentially target ASIC3. Chapter 1 reviews the current state of knowledge on venom derived ASIC modulators, with a particular focus on their molecular interaction with ASICs, what they have taught us about channel structure, and what they may still reveal about ASIC function and pathophysiological roles.

Venom peptides are often disulfide bonded making their recombinant expression challenging. Chapter 2 describes a periplasmic Escherichia coli protocol for production of correctly folded peptides that was used throughout this thesis. Using the two-electrode voltage-clamp technique, Chapter 3 shows that a wide variety of voltage- and ligand-gated ion channels have the same channel properties and pharmacological profiles when expressed in either Xenopus laevis or X. borealis oocytes. This voltage-clamp technique was heavily used throughout this thesis to perform functional studies of venom peptides.

The spider venom peptide PcTx1 is the best studied ASIC modulator; it has an IC50 of ~1 nM at rat ASIC1a and is neuroprotective in rodent models of ischemic stroke. Chapter 4 examines the molecular interaction between PcTx1 and both human ASIC1a and the off- target ASIC1b subtype, where little experimental work has been done. We show that although PcTx1 is 10-fold less potent at human ASIC1a than the rat channel, the apparent affinity for the two channels is comparable. The pharmacophore of PcTx1 for human ASIC1a and rat ASIC1b was examined via alanine scanning mutagenesis and uncovered residues that show subtle ASIC1 species and subtype-dependent differences in activity that may allow for further manipulation to develop more selective PcTx1 analogues.

The ASIC3 inhibitor APETx2 is analgesic in rodent models of chemically-induced, inflammatory and postoperative pain, providing strong evidence for the ASIC3 subtype as a potential pain target. Despite the comprehensive structure-activity studies of APETx2, the mechanism of action and channel binding site have remained elusive. Chapter 5 fills this

II gap in knowledge of an important ASIC research tool and also reveals subtle differences in the pharmacophore of APETx2 for inhibition of ASIC3 and potentiation of ASIC1b. We propose that APETx2 binds to a novel region for peptide modulators of ASICs and acts by preventing the closed-to-open gating transition of rat ASIC3.

Mambalgins are a group of three-finger toxins isolated from black and green mamba snake venoms that target ASIC1a and ASIC1b containing channels. Unique among ASIC modulators, the potent inhibitory activity of mambalgins at rodent ASIC1b was crucial in demonstrating that ASIC1b, and not ASIC1a, is important for peripheral pain sensing in rodents. Chapter 6 shows that the efficacy of mambalgins varies between the ASIC1 splice variants ASIC1a and ASIC1b, in both human and rat channels. Strikingly, mambalgin is a potentiator of human ASIC1b under certain conditions. Furthermore, these pharmacological differences are due to both the molecular interactions at the binding sites and the different ways mambalgin can modify the gating characteristics of specific ASIC variants.

Chapters 7 describes the isolation, pharmacological characterisation and chemical stability of the novel spider venom peptide Hm3a from the venom of the tarantula Heteroscodra maculata. Hm3a is a close homolog of PcTx1 with five amino acid substitutions and a three residue C-terminal truncation. Despite its high sequence similarity with PcTx1 and similar pharmacology, Hm3a showed higher levels of stability over 48 h and will be particularly useful when stability in biological fluids is required, for example in long term in vitro cell- based assays and in vivo experiments.

Chapter 8 describes the characterisation of conorfamides As1a and As2a from the venom of the cone snail Conus austini. Amidated peptide variants altered desensitization of ASIC1a and 3, and a lysine to arginine mutation introduced ASIC1a peak current potentiation. These conorfamides also inhibited a7 and muscle-type nicotinic acetylcholine receptors (nAChR) at nanomolar concentrations. These are the first conorfamides with the dual pharmacology described to date.

III Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co- authors for any jointly authored works included in the thesis.

IV Publications included in this thesis

1. B Cristofori-Armstrong# and LD Rash# (2017) Acid-sensing ion channel (ASIC) structure and function: Insights from spider, snake and sea anemone. Neuropharmacology 127: 173–184. 2. SY Er*, B Cristofori-Armstrong*, P Escoubas and LD Rash (2017) Discovery and molecular interaction studies of a highly stable, tarantula peptide modulator of acid- sensing ion channel 1. Neuropharmacology 127: 185–195. 3. B Cristofori-Armstrong, MS Soh, S Talwar, DL Brown, JDO Griffin, Z Dekan, JL Stow, GF King, JW Lynch and LD Rash (2015) Xenopus borealis as an alternative source of oocytes for biophysical and pharmacological studies of neuronal ion channels. Scientific Reports 5: 14763.

V Submitted manuscripts included in this thesis

1. B Cristofori-Armstrong*, NJ Saez*, IR Chassagnon, GF King, and LD Rash. The modulation of acid-sensing ion channel 1 by PcTx1 is pH-, subtype- and species- dependent: importance of interactions at the channel subunit interface and potential for engineering selective analogues. Biochemical Pharmacology. Revisions resubmitted.

2. A-H Jin*, B Cristofori-Armstrong*, LD Rash, SAR González, RJ Lewis, PF Alewood, and I Vetter. Novel conorfamides from Conus austini venom modulate both nicotinic acetylecholine receptors and acid-sensing ion channels. Biochemical Pharmacology. Submitted.

Other publications during candidature

Book Chapter

1. NJ Saez, B Cristofori-Armstrong, R Anangi, and GF King (2017) A Strategy for Production of Correctly Folded Disulfide-rich Peptides in the Periplasm of E. coli. Methods in Molecular Biology 1586: 155–180.

Journal Article (* joint first author; # co-corresponding author)

1. A Silva*, B Cristofori-Armstrong*, LD Rash, WC Hodgson and GK Isbister (2018) Defining the role of post-synaptic -neurotoxins in paralysis due to envenoming in humans. Cellular and Molecular Life Sciences 75: 4465–4478

2. H Shen, Z Li, Y Jiang, X Pan, J Wu, B Cristofori-Armstrong, JJ Smith, YKY Chin, J Lei, Q Zhou, GF King, N Yan (2018) Structural basis for the modulation of voltage-gated sodium channels by animal toxins. Science 362: eaau2596

3. B Madio, S Peigneur, YKY Chin, BR Hamilton, ST Henriques, JJ Smith, B Cristofori- Armstrong, Z Dekan, BA Boughton, PF Alewood, J Tytgat, GF King and EAB Undheim (2018) PHAB toxins: a unique family of sea anemone toxins evolving via intra-gene concerted evolution defines a new peptide fold. Cellular and Molecular Life Science 75: 4511–4524.

4. JYP Lee*, NJ Saez*, B Cristofori-Armstrong*, R Anangi, GF King, MT Smith and LD Rash (2018) Inhibition of acid-sensing ion channels by diminaene and APETx2 evoke

VI partial and highly variable antihyperalgesia in a rat model of inflammatory pain. British Journal of Pharmacology 175: 2204–2218.

5. SS Pineda, PA Chaumeil, A Kunert, Q Kaas, MWC Thang, L Li, M Nuhm, V Herzig, NJ Saez, B Cristofori-Armstrong, R Anangi, S Senff, D Gorse and GF King (2017) ArachnoServer 3.0: an online resource for automated discovery, analysis and annotation of spider toxins. Bioinformatics 34: 1074–1076.

6. JS Wingerd, CA Mozar, CA Ussing, SS Murali, YKY Chin, B Cristofori-Armstrong, T Durek, J Gilchrist, CW Vaughan, F Bosman, DJ Adams, RJ Lewis, PF Alewood, M Mobli, MJ Christie and LD Rash (2017) The tarantula toxin /-TRTX-Pre1a highlights the importance of the S1-S2 voltage-sensor region for sodium channel subtype selectivity. Scientific Reports 7: 974.

7. NJ Saez*, E Deplazes*, B Cristofori-Armstrong*, IR Chassagnon, X Lin, M Mobli, LD Rash and GF King (2015) Molecular dynamics and functional studies define a hot spot of crystal contacts essential for PcTx1 inhibition of acid-sensing ion channel 1a. British Journal of Pharmacology 172: 4985–4995.

8. CY Chow, B Cristofori-Armstrong, EAB Undheim, GF King and LD Rash (2015) Three peptide modulators of the human voltage-gated sodium channel 1.7, an important analgesic target, from the venom of an Australian Tarantula. Toxins 7: 2494–2513.

Conference Abstract

B Cristofori-Armstrong, JJ Smith, K Voll, S Reynaud, M Mobli, LD Rash (2018) Functional studies of the pain target acid-sensing ion channel 1. Ion Channels Gordon Research Conference. July 2018 (Poster Presentation).

B Cristofori-Armstrong, A Silva, WC Hodgson and LD Rash (2017) Defining the role of post-synaptic -neurotoxins in paralysis due to envenoming in humans. International Postgraduate Symposium in Biomedical Science. November 2017 (Poster presentation).

B Cristofori-Armstrong, JYP Lee, NJ Saez, R Anangi, GF King, MT Smith and LD Rash (2017) Understanding the interaction of the peptide APETx2 with acid-sensing ion channels, and it’s biphasic analgesic properties in rodent inflammatory pain. Gage Conference Ion Channels and Transporters. April 2017 (Oral Presentation).

VII Contributions by others to the thesis

Dr Lachlan Rash (primary supervisor) and Prof Glenn King (associate supervisor) provided advice on project design and coordination throughout my thesis.

Chapter 3: Scanning electron microscopy was performed with Darren Brown, and oocyte images with John Griffin. Prof Joseph Lynch and Dr Rash contributed to editing.

Chapter 4: Comparison of PcTx1 mechanism of action at rat and human channels were performed in collaboration with Dr Rash. Dr Rash also contributed to editing.

Chapter 5: APETx2 mechanism of action studies were performed in collaboration with Dr Rash. The inflammatory pain animal study was performed by Dr Jia Yu (Peppermint) Lee. Dr Rash also contributed to editing.

Chapter 6: Dr Thomas Durek synthesised mambalgin-2 and Kirsten Voll isolated native mambalgin-3. Dr Rash contributed to editing.

Chapter 7: This was a collaborative project with Dr Sing Yan (Jessie) Er, and data was collected together. Dr Rash contributed to editing.

Chapter 8: Conus austini venom was collected by Mr Sergio González, and peptides isolated together with A/Prof Irina Vetter. Dr Ai-Hua (Jean) Jin synthesised peptides and contributed to conception. A/Prof Irina Vetter also contributed to conception and editing.

VIII Statement of parts of the thesis submitted to qualify for the award of another degree

No works submitted towards another degree have been included in this thesis.

Research Involving Human or Animal Subjects

The protocol for Xenopus laevis and X. borealis studies was approved by the Anatomical Biosciences Animal Ethics Committee at The University of Queensland (Approval Number: QBI/059/13/ARC/NHMRC and QBI/AIBN/087/16/NHMRC/ARC). Copies of each ethics approval letter are included in Appendix I. No human subjects were involved in this research.

IX Acknowledgements

I would like to express thanks to my supervisors Lachlan Rash and Glenn King for all the support and freedom they gave me during my PhD. Lachlan has been a great teacher and friend both in and out of the lab, and Glenn has taught me invaluable lessons for moving forward in research. I would not be in the place I am now without both their help.

I would also like to thank my committee members Rohan Teasdale, Tina Schroeder, and Johan Rosengren for always being available to listen and provide honest feedback on my lab work and career thoughts. A special thanks also goes to Amanda Carozzi, Cody Mudgway, and Olga Chaourova (the IMB postgrad team) for their propensity of always being there when I’d hoped and needed some help with matters outside the lab.

Furthermore, I’d like to thank all members past and present from the King and Rash labs, members of the Lynch lab and Aquatics Team of UQBR, non-research staff that help the IMB run every day, members of the Clark and Rosengren labs at SBMS, Bruno Madio and Elena Budusan for the coffee breaks and chats, and Mehdi Mobli and Eivind Undheim for all their bits of knowledge that have helped me along the way. It’s been a pleasure working with everyone and I hope it continues in the future.

A big thank you to Jen for always being there.

Financial support

This research was supported by an Australian Government Research Training Program Scholarship.

Keywords

Acid-sensing ion channel, venom peptide, recombinant expression, two-electrode voltage clamp, mechanism of action, structure-activity relationship, ion channel, pharmacology

X Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 060110, Receptors and Membrane Biology, 60%

ANZSRC code: 111501, Basic Pharmacology, 30%

ANZSRC code: 060199, Biochemistry and Cell Biology not elsewhere classified, 10%

Fields of Research (FoR) Classification

FoR code: 0601, Biochemistry and Cell Biology, 70%

FoR code: 1115, Pharmacology and Pharmaceutical Sciences, 30%

XI Table of Contents

Abstract ...... II

Declaration by author ...... IV

Publications included in this thesis...... V

Submitted manuscripts included in this thesis ...... VI

Other publications during candidature ...... VI

Contributions by others to the thesis ...... VIII

Statement of parts of the thesis submitted to qualify for the award of another degree ...... IX

Research Involving Human or Animal Subjects ...... IX

Acknowledgements ...... X

Financial support ...... X

Keywords ...... X

Australian and New Zealand Standard Research Classifications (ANZSRC) ...... XI

Fields of Research (FoR) Classification ...... XI

List of figures ...... XVII

List of tables ...... XIX

List of abbreviations used in the thesis ...... XX

CHAPTER 1 ...... 1

Acid-sensing ion channel (ASIC) structure and function: insights from spider, snake, sea anemone, and cone snail venoms ...... 1

1.1 INTRODUCTION TO ACID-SENSING ION CHANNELS ...... 2

1.2 VENOM PEPTIDES TARGETTING ASICS ...... 6

1.2.1 PcTx1 ...... 7

1.2.2 Hm3a ...... 13

1.2.3 Hi1a ...... 14

1.2.4 APETx2 ...... 15

XII 1.2.5 MitTx ...... 16

1.2.6 Mambalgins ...... 18

1.2.7 Ugr 9-1 ...... 21

1.2.8 PhcrTx1 ...... 22

1.2.9 -Dendrotoxin ...... 22

1.2.10 Conorfamide RPRFa (also CNF-Tx1.1) ...... 23

1.3 CONCLUSIONS ...... 23

1.4 THESIS AIMS ...... 24

CHAPTER 2 ...... 25

Materials and Methods ...... 25

2.1 RECOMBINANT PEPTIDE PRODUCTION ...... 26

2.1.1 Plasmid design ...... 27

2.1.2 Expression and purification ...... 27

2.1.3 HPLC purification ...... 28

2.1.4 MALDI-TOF mass spectrometry ...... 29

2.2 OOCYTE ELECTROPHYSIOLOGY ...... 29

2.2.1 Two-electrode voltage clamp recordings ...... 29

2.2.2 Ethics statement ...... 30

2.3 MOLECULAR BIOLOGY ...... 30

2.3.1 Mutagenesis ...... 30

CHAPTER 3 ...... 32

Xenopus borealis as an alternative source of oocytes for biophysical and pharmacological studies of neuronal ion channels ...... 32

3.1 INTRODUCTION ...... 33

3.2 METHODS ...... 34

3.2.1 Peptide preparation ...... 34

3.2.2 Data analyses ...... 34

3.2.3 Scanning electron microscopy ...... 34

3.3 RESULTS ...... 35 XIII 3.3.1 Comparison of X. laevis and X. borealis oocytes ...... 35

3.3.2 Voltage-gated ion channels ...... 36

3.3.3 Ligand-gated ion channels ...... 38

3.3.4 Scanning electron microscopy ...... 39

3.4 DISCUSSION ...... 41

2.5 SUPPLEMENTARY INFORMATION ...... 43

CHAPTER 4 ...... 44

Defining the molecular interactions and mechanism of PcTx1 activity at rat and human ASIC1a and rat ASIC1b ...... 44

4.1 INTRODUCTION ...... 45

4.2 METHODS ...... 47

4.2.1 Data analyses ...... 47

4.3 RESULTS ...... 47

4.3.1 PcTx1 is less potent against human ASIC1a than the rat isoform ...... 47

4.3.2 Pharmacophore of PcTx1 against human ASIC1a ...... 51

4.3.3 Pharmacophore of PcTx1 at rat ASIC1b ...... 53

4.4 DISCUSSION ...... 56

CHAPTER 5 ...... 62

APETx2 mechanism of action, binding site, and rat ASIC1b pharmacophore...... 62

5.1 INTRODUCTION ...... 63

5.2 MATERIALS AND METHODS ...... 64

5.2.1 Data analyses ...... 64

5.2.2 Peptides ...... 64

5.3 RESULTS ...... 64

5.3.1 APETx2 mechanism of action ...... 64

5.3.2 ASIC3 mutagenesis for the APETx2 binding site ...... 66

5.3.3 Functional competition assays of APETx2 ...... 67

5.3.4 ASIC1a:1b chimeras to uncover the APETx2 binding site ...... 69

5.3.5 APETx2:rASIC1b pharmacophore ...... 70 XIV 5.4 DISCUSSION ...... 71

CHAPTER 6 ...... 73

ASIC1 subtype and species dependent activity of mambalgins: not a potent inhibitor of human ASIC1b ...... 73

6.1 INTRODUCTION ...... 74

6.2 METHODS ...... 75

6.2.1 Peptides ...... 75

6.2.2 Data analyses ...... 76

6.3 RESULTS ...... 76

6.3.1 Ma-3 subtype selectivity and mechanism of action ...... 76

6.3.2 Ma-3 pharmacophore ...... 80

6.3.3 Ma-3 binding site ...... 82

6.4 DISCUSSION ...... 85

CHAPTER 7 ...... 93

Discovery and molecular interaction studies of a highly stable, tarantula peptide modulator of acid-sensing ion channel 1 ...... 93

7.1 INTRODUCTION ...... 94

7.2 MATERIALS AND METHODS ...... 95

7.2.1 Venom peptide purification and characterisation ...... 95

7.2.2 Peptide sequencing ...... 96

7.2.3 Data analyses ...... 96

7.2.4 Stability assays ...... 96

7.2.5 Deposition of protein sequence information ...... 97

7.3. RESULTS AND DISCUSSION ...... 97

7.3.1 Identification and purification of Hm3a ...... 97

7.3.2 Recombinant production of Hm3a ...... 99

7.3.3 Effects of Hm3a on ASICs ...... 100

7.3.4 Mechanism of action of Hm3a on ASIC1 ...... 103

XV 7.3.5 Characterizing the molecular interactions involved in subtype and species selectivity of Hm3a on ASIC1 ...... 105

7.3.6 Stability of Hm3a and PcTx1 ...... 109

7.5. CONCLUSIONS ...... 111

7.6 SUPPLEMENTARY INFORMATION ...... 112

CHAPTER 8 ...... 115

Novel conorfamides from Conus austini venom modulate both nicotinic acetylcholine receptors and acid-sensing ion channels ...... 115

8.1 INTRODUCTION ...... 116

8.2 METHODS ...... 117

8.2.1 Crude venom extraction and fractionation ...... 117

8.2.2 FLIPR-based screen to isolate conopeptides ...... 118

8.2.3 Chemistry-based sequencing ...... 118

8.2.4 Peptide synthesis and co-elution ...... 118

8.2.5 Data analyses ...... 119

8.3 RESULTS ...... 119

8.3.1 Peptide isolation and sequencing from C. austini venom ...... 119

8.3.2 Peptide synthesis and co-elution ...... 121

8.3.3 Pharmacology at ASICs ...... 122

8.3.4 Pharmacology at nAChR ...... 123

8.4 DISCUSSION AND CONCLUSIONS ...... 124

8.5 SUPPLEMENTARY INFORMATION ...... 126

CHAPTER 9 ...... 127

Conclusions and future directions ...... 127

REFERENCES ...... 132

APPENDIX I ...... 153

XVI List of figures

Figure 1.1: ASIC gating conformations, domain organisation, and venom peptide binding sites and functional summary ...... 5 Figure 1.2: Venom peptide modulators of ASICs ...... 9 Figure 1.3: Summary of the active surface of ASIC targeting peptides derived from crystallography and mutagenesis data ...... 11 Figure 2.1: Schematic of the pathway for production of secreted proteins in E. coli ...... 26 Figure 2.2: Expression of DRPs ...... 28 Figure 3.1: Comparison of defolliculated stage V–VI oocytes from X. laevis and X. borealis ...... 35 Figure 3.2: Endogenous Ca2+-activated chloride currents in naïve X. laevis and X. borealis oocytes ...... 36

Figure 3.3: Voltage-dependent properties of KV and NaV channels expressed in X. laevis and X. borealis oocytes ...... 37 Figure 3.4: The effect of activating and antagonist ligands on different ASIC subtypes expressed in X. laevis and X. borealis oocytes ...... 38 Figure 3.5: SEM images of defolliculated X. laevis and X. borealis stage V–VI oocytes prior to and following protease treatment ...... 40 Figure 4.1: Comparison of PcTx1 activity at rat and human ASIC1a ...... 48 Figure 4.2: pH-dependent effects of PcTx1 for ASIC1a ...... 50 Figure 4.3: Activity of PcTx1 at different conditioning pHs ...... 51 Figure 4.4: Structure-activity relationship of PcTx1 at hASIC1a ...... 53 Figure 4.5: Comparative activity of PcTx1 mutants at rASIC1a and rASIC1b ...... 55 Figure 4.6: Subtype specificity of PcTx1 analogues ...... 56 Figure 5.1: APETx2 mechanism of action at rat ASIC3 ...... 66 Figure 5.2: APETx2 binding site via ASIC3 mutagenesis ...... 67 Figure 5.3: Functional competition studies of APETx2 with PcTx1 at rASIC1b ...... 68 Figure 5.4: Functional competition studies of APETx2 with MitTx at rASIC3...... 69 Figure 5.5: ASIC1a:1b chimeras for APETx2 binding site ...... 70 Figure 5.6: APETx2 interacts with both rat ASIC3 and ASIC1b ...... 71 Figure 6.1: Production of recombinant Ma-3 ...... 75 Figure 6.2: Mambalgin subtype selectivity ...... 77 Figure 6.3: Ma-3 mechanism of action ...... 78 Figure 6.4: Ma-3 calcium-dependence and kinetics of activity at rat ASIC1a ...... 79 Figure 6.5: Activity of Ma-3 mutants at rat ASIC1a...... 80 Figure 6.6: Activity of Ma-3 mutants at rat ASIC1b ...... 81 Figure 6.8: Binding site interactions that determine the potency difference of Ma-3 at rat and human ASIC1a ...... 83 XVII Figure 6.9: Amino acids that differ between rat ASIC1a and ASIC1b in the Ma-3 binding site...... 84 Figure 6.10: ASIC mutations in the wider Ma-3 binding site ...... 85 Figure 6.11: Comparison of published binding models of mambalgin to ASIC ...... 89 Figure 6.12: Summary of published findings and data presented here of the Ma-3 interaction with ASIC1a ...... 92 Figure 7.1: Isolation of Hm3a from H. maculata venom ...... 98 Figure 7.2: Production of recombinant Hm3a ...... 99 Figure 7.3: Concentration-effect curves of recombinant Hm3a for homomeric rASIC1a, rASIC2a, and rASIC3 ...... 101 Figure 7.4: Mechanism of action of Hm3a on rASIC1 ...... 104 Figure 7.5: Molecular interactions involved in species and subtype selectivity ...... 108 Figure 7.6: Comparative stability studies of Hm3a and PcTx1 ...... 110 Figure S7.1: Example current traces from Xenopus oocyte data corresponding to concentration- response curves presented in Figures 7.1D and 7.3 ...... 112 Figure S7.2: Example current traces from Xenopus oocyte data corresponding to on- and off-rates presented in Figures 7.4D and E ...... 113 Figure S7.3: Potentiation of rASIC1a currents by Hm3a with less than 10 s peptide exposure .. 113 Figure S7.4: pH-dependence of steady-state desensitization and activation for wild-type and mutant rASIC1a channels used in this study ...... 114 Figure S7.5: Equivalent data set from Figure 7.5B showing the concentration-effect curves of Hm3a on the channel mutant rASIC1a_F350A ...... 114 Figure 8.1: Assay guided fractionation of Conus austini venom ...... 120 Figure 8.2: Peptide sequence determination ...... 121 Figure 8.3: Synthesis of peptides and co-elution with native peptides for Conorfamide_As1a and Conopeptide_As1b...... 121 Figure 8.4: ASIC activity ...... 122 Figure 8.5: nAChR activity ...... 123 Figure 8.6: Sequence alignment of the conorfamide family. Background colouring indicates residue conservation...... 124 Figure S8.1: Full data set for peptides (50 M) screened across rat ASIC subtypes ...... 126

XVIII List of tables

Table 1.1: Activity summary of ASIC modulators isolated from animal venoms ...... 6 Table 2.1: Disulfide-rich venom peptides expressed for this thesis ...... 26 Table 3.1: Maximal current amplitudes recorded from exogenous channels expressed in X. laevis and X. borealis oocytes ...... 37 Table S3.1: Comparison of half-maximal response and slope values for ion channels tested in X. laevis and X. borealis oocytes and references to studies reporting similar values for these parameters ...... 43 Table 4.1: Effect of increasing concentrations of PcTx1 on steady-state desensitisation ...... 49

Table 4.2: Summary of wild-type and mutant PcTx1 IC50 values on rat and human ASIC1a ...... 52

XIX List of abbreviations used in the thesis

Act activation ASIC acid-sensing ion channel BSA bovine serum albumin cASIC1a chicken ASIC1a CHCA -cyano-4-hydroxycinnamic acid CNS central nervous system DRG dorsal root ganglion DRP disulfide-rich peptide ECD extracellular domain FaNaC FMRF-amide-gated sodium channel FMRFa Phe-Met-Arg-Phe-amide (Phe-Met-Arg-Phe-NH2) IPTG isopropyl -D-1-thiogalactopyranoside hASIC human ASIC His6 hexahistidine (His6)-tag LB Luria-Bertani Ma mambalgin MALDI-TOF matrix-assisted laser desorption/ionisation time-of-flight MBP maltose-binding protein m/z mass-to-charge ratio nAChR nicotinic acetylecholine receptor PcTx1 psalmotoxin-1 PNS peripheral nervous system SAR structure-activity relationship rASIC rat ASIC SDS-PAGE sodium dodecyl sulphate polyacrylamide gel electrophoresis SEM scanning electron microscopy SSD steady-state desensitisation SPPS solid-phase peptide synthesis TEV tobacco etch virus TFA trifluoroacetic acid TMD transmembrane domain RP-HPLC reverse-phase high performance liquid chromatography

XX

CHAPTER 1

Acid-sensing ion channel (ASIC) structure and function: insights from spider, snake, sea anemone, and cone snail venoms

1 1.1 INTRODUCTION TO ACID-SENSING ION CHANNELS

Acid-sensing ion channels (ASICs) were described at the sequence level in 1996 (1-3) and shortly thereafter identified as proton-gated cation channels (4). However some 15 years before these discoveries, Krishtal and Pidoplichko reported the first instances of a proton- induced conductance in sensory neurons (5,6). Since then it has been shown that ASICs are abundantly expressed throughout neuronal and non-neuronal tissues, where they are primary acid-sensors. Under physiological conditions, pH is tightly regulated and an extracellular pH of ~7.4 is crucial for normal cell function. Acidosis of sufficient magnitude to activate ASICs occurs in many pathological conditions exemplified by inflammation, ischemia, tumours and trauma (7-9).

Distribution of ASICs ASICs are a subfamily within the broader superfamily of Na+ conducting channels that includes the epithelial Na+ channel (ENaC), the degenerins (DEG), and the invertebrate FMRF-amide-activated channel (FaNaC). There are four human ASIC (hASIC) genes (ASIC1–4, renamed from the previous ACCN designation) encoding at least six major subunits (ASIC1a, ASIC1b, ASIC2a, ASIC2b, ASIC3, and ASIC4). The ASIC1 and ASIC2 genes are alternatively spliced to produce “a” and “b” isoforms, which differ in the N-terminal third of the protein (10,11). Additionally, there are three functional splice variants of hASIC3 (a, b and c), which do not appear to be present in rodents (12).

ASIC1a, ASIC2a, and ASIC4 are highly expressed in the brain and spinal cord, as well as to lesser degrees in peripheral neurons (13). Within the central nervous system these isoforms have also been implicated in diverse physiological roles and functions including fear conditioning, synaptic plasticity, learning and memory (7,14,15). ASIC3 was considered to be restricted in expression to the peripheral nervous system where it is highly expressed and involved in pain sensation (16), however subsequent work has shown expression in the brain where its function is less clear (17). The ASIC1b subtype is expressed exclusively in the peripheral sensory neurons where it also plays a role pain perception (18).

Biophysical properties of ASICs Three ASIC subunits combine to form functional homo- and heterotrimers (19) with different biophysical properties (20-22). The rapid, transient peak-current of ASICs is selective for Na+ over K+ by a factor of ~10, and desensitises within seconds in the continued presence of an acidic pH. That is, ion channels enter a non-conducting state and fail to respond to

2 additional acidification (4). ASIC activation is pH-dependent (i.e., current amplitude increases with decreasing pH until a maximal current amplitude is reached) (23). The conditioning solution is typically the solution in which ASICs are exposed to during electrophysiology assays before applying a low pH stimulus. Prolonged exposure of ASICs to mildly acidic pH in the conditioning solution or slow acidification results in steady-state desensitisation (SSD), which is where the channel enters a desensitised state without being preceded by an observable open conducting state (24). This has been proposed to be due to the rapid kinetics of desensitisation and a slow on-rate for protons at lower concentrations, thereby masking any transition through an open state (21). This model proposed by Gründer and Pusch is however qualitative in nature, and it was noted that the possibility of a simple closed to desensitised transition cannot be excluded. In this alternative, and possibly simplest explanation, the gating of ASICs does allow a direct transition from the closed to the desensitised state, as desensitisation is more pH-sensitive than activation. In this case, upon slow or mild acidification, the channels would desensitise before activation could take place (thus any observable current). Whereas during rapid acidification, channel activation is observable due to the faster kinetics of activation compared to desensitisation (however channel opening would still be inherently influenced by the rate of desensitisation and likely be underestimated). Nevertheless, the complete picture of ASIC gating remains to be solved and remains an area of intense research interest.

A massive leap in our understanding of the structure of ASICs occurred when the crystal structure of chicken ASIC1 (ASIC1a sequence homologue) was published in 2007 (19), even though the construct used for crystallisation was a non-functional channel with heavily truncated N- and C-termini. The structure revealed the trimeric nature of ASICs, and showed that each subunit consists of short intracellular N- and C-termini and two hydrophobic transmembrane (TM) domains separated by a large extracellular domain of ~370 residues. The structure of an individual subunit was described as resembling a hand clenched around a ball comprising a wrist, palm, finger, knuckle, thumb, and β-ball domains (see Figure 1.1A). This structure also demonstrated that the N-terminal third that differs between spilce isoforms is not an isolated domain, and makes contact with several other regions of the channel including the C-terminal transmembrane domain and functionally important thumb domain at the subunit interface. Subsequently a less severely N-terminally truncated and functional cASIC1 construct, named 13 cASIC1, was crystallised in the presumed desensitised state and differed from the previous structure primarily in the orientation of the TM domains (25). 3 After the initial crystallisation of apo structures, subsequent cASIC1:venom peptide co- crystal structures have been solved in different states and provided further insights into ASIC structure and gating (Figure 1.1C and D) (26,27). Currently it is thought that the MitTx-bound structure most faithfully represents the true open state (Figure 1.1D) (26). Importantly, the desensitised and open state structures are extremely similar in conformation in the extracellular region where peptide modulators have been shown to bind. Recently the resting state structure of cASIC1 was solved without any ligands bound, and revealed significant conformational rearrangements in the extracellular domain compared to the open and desensitised states (Figure 1.1B) (28). The most significant differences in the closed state were observed in the finger and thumb domain, that are further away from the centre of the channel and result in an expanded acidic pocket, while the transmembrane domains present a closed pore. As the channel is exposed to a drop in pH, residues in the acidic pocket are thought to be protonated resulting in closure of the thumb domain towards the three-fold molecular axis (28-30). This movement is conferred to the palm domain that ultimately results in expansion of the fenestrations leading to the transmembrane domain pore, and allows passage of ions that are observed as conductance. After activation, desensitisation occurs almost immediately while at low pH. A rearrangement of the 11-12 linker in the palm domain uncouples the extracellular domain from the lower channel and results in a relaxing of the transmembrane domains that closes the pore and prevents ion conduction. The final rearrangement back to the closed state comes about as residues in the extracellular domain become deprotonated and carboxyl-carboxylate pairs move away from each other as the acidic pocket is expanded and ready again for protonation (25,26,28).

Despite the growing interest in ASICs, and the increasing volume of work being produced in the field, there are still many unanswered questions. Venom derived compounds have greatly enhanced our understanding of ASICs to date, and are likely to play an important role in the future. Here, we will focus on our current understanding of these venom-derived compounds, how they have provided critical information on ASIC structure and function, and why this knowledge is important for future studies using venom peptides to elucidate ASIC function and pharmacology. Please refer to these reviews for more in-depth coverage of recent advances in genetic models (13) and biophysical properties (21) of ASICs, and more comprehensive coverage of the in vivo use of ASIC modulating venom peptides to study the pathological roles of ASICs (31,32).

4

Figure 1.1: ASIC gating conformations, domain organisation, and venom peptide binding sites and functional summary. (A) cASIC1 structure in the apo, desensitised state without any peptide bound (PDB code 4NYK; (25)), with domains of a single subunit within the trimer colour coded. (B) The closed state of cASIC1 was recently solved by both crystallography and cryo-EM (PDB code 5WKU; (28)). This structure showed significant rearrangements in the thumb and finger domains away from central axis compared to the desensitised and open states, resulting in an expanded acidic pocket. (C) PcTx1 bound to cASIC1 at low pH (PDB code 4FZ1; (27)), and (D) MitTx bound to cASIC1 (PDB code 4NTW; (26)), where both structures were reported at the time of publication to be in the open state. However, there are clear differences in the TM region and it is now thought the MitTx most faithfully represents the Na+ conductive, proton-gated open state of ASICs. ASIC modulators are shown next to the binding state each peptide favours, for the subtype shown in brackets.  in red indicates inhibitors,  in blue potentiators of pH-gated currents, and  in green direct agonists of the ASIC subtype in brackets for each ligand. Peptide binding sites are in bold, and the grey lines depict the TM region based on the hydrophobicity of the protein surface. Note the functional information listed for each ligand is not complete and is a simplified summary; please refer to the text and Table 1.1 for more details. 5 1.2 VENOM PEPTIDES TARGETTING ASICS

Many endogenous and exogenous modulators of ASICs have now been discovered and range from small molecules to larger proteins (33). Although the lack of truly selective modulators of ASICs has made it difficult to tease out the relative importance of specific ASIC subtypes, the most interesting and selective tools available to date have all come from animal venoms.

Table 1.1: Activity summary of ASIC modulators isolated from animal venoms. IC50 and EC50 values are for rat isoforms unless stated otherwise. Activity is colour coded into red, blue, and green indicating inhibition , potentiation , and agonism , respectively.

Peptide Venom source Activity PcTx1 Spider (Psalmopoeus  ASIC1a ~1 and ~3 nM (rat and human) cambridgei)  ASIC1a/2a ~35–85% at 50 nM  ASIC1a/2b ~3 nM  ASIC1b ~100 nM  cASIC1 ~10–189 nM  cASIC1 ~135% at 30 nM Hm3a Spider (Heteroscodra  ASIC1a 1.3 and 39.7 nM (rat and human) maculata)  ASIC1b 46.5 and 178 nM (rat and human)  ASIC1a/1b 17.4 nM APETx2 Sea anemone  ASIC3 ~60 and 175 nM (rat and human) (Anthopleura elegantissima)  ASIC3/1a 2 M  ASIC3/1b 900 nM  ASIC3/2b 117 nM Hcr 1b-1 Sea anemone  ASIC3 5.5 M (human) (Heteractis crispa) MitTx Snake (Micrurus tener  ASIC1a 9 nM tener)  ASIC1b 23 nM  ASIC2a 75 nM  ASIC3 830 nM  cASIC1 ~15% at 300 nM Ma-1 Snake (Dendroaspis  ASIC1a 11–55 and 127 nM (rat and human) polylepis polylepis, Ma-2  ASIC1b 44–192 nM Dendroaspis Ma-3 augusticeps)  ASIC1a/1b 72 nM  ASIC1a/2a 252 nM

6  ASIC1a/2b 61 nM Ugr 9-1 Sea anemone (Urticina  ASIC3 10 M (human, peak) grebelnyi)  ASIC3 1.5 M (human, sustained) PhcrTx1 Sea anemone  100 nM (rat sensory neurons) (Phymanthus crucifer) -DTx Snake (Dendroaspis  800 nM (rat sensory neurons) augusticeps)

RPRFa Cone snail (Conus  ASIC3 4.2 M (I5s/Ipeak) textile)

1.2.1 PcTx1

The prototypical, selective ASIC inhibitor PcTx1 (also known as π-TRTX-Pc1a) is a 40-residue peptide isolated from the venom of the Trinidad chevron tarantula Psalmopoeus cambridgei (Figure 1.2A) (34). Although isolated as the first highly potent and selective inhibitor of homomeric ASIC1a (rat IC50 ~1 nM conditioned at pH 7.45 and human IC50 ~3 nM conditioned at pH 7.2), PcTx1 has subsequently been shown to also inhibit currents in ASIC1a/2b co-expressing cells (mouse

IC50 ~3 nM) and heteromeric rat ASIC1a/2a channels (by ~35–85% at 50 nM conditioned at pH 6.95 depending on trimer composition) (35,36). At higher concentrations PcTx1 potentiates ASIC1b (rat EC50 ~100 nM) (37) and is an agonist and/or potentiator of cASIC1 depending on the conditions used (EC50 10–189 nM) (38,39). The NMR structure of PcTx1 (Figure 1.2B), which was first solved in 2003 and further refined in 2011, revealed the presence of an inhibitor cysteine knot (ICK) motif and that the dominant -hairpin loop, which presents some of the key pharmacophore residues, is highly flexible (40,41).

7

8 Figure 1.2: Venom peptide modulators of ASICs. (A) Amino acid sequence of venom peptides grouped with homologues where present, and differences in sequences are indicated by grey boxes. Peptide residues interacting with the channel are shown by triangles, with red/orange based on functional mutagenesis data and blue from crystal contacts. (B) Three dimensional structure of PcTx1 (PDB code 2KNI;(41)), APETx2 (PDB code 2MUB; (42)), Hi1a (PDB Code 2N8F; (43)), Ma- 1 (PDB codes 5DU1, 5DZ5; (44)), Ma-2 (PDB code 2MFA; (45)), MitTx (PDB code 4NTW; (26)), Ugr 9-1 (PDB code 2LZO; (46)), and -DTx (PDB code 1DTX; (47)). All peptides are shown in the same scale and disulfide bonds shown in red. Structures are not available for all peptides and have only been shown where available.

PcTx1 binds in a state-dependent manner to both ASIC1a and ASIC1b. It acts as a gating modifier of ASIC1a by increasing the channel's apparent affinity for protons, thereby rendering the channel inactive at resting physiological pH of 7.4 (48). This effect can be seen as a shift in the steady-state desensitisation (SSD) and activation curves to less acidic pH values. It is in fact this mechanism of action that renders PcTx1 less potent at human as compared to rodent ASIC1a (49). The SSD of hASIC1a is inherently less sensitive to pH, thus a higher concentration of PcTx1 is required to capture the channel in a desensitised (and non-conducting) state when conditioned at pH 7.4. In agreement with this, mutation of five residues of hASIC1a to the equivalent mouse ASIC1a (mASIC1a) residues makes the pH sensitivity of the species variants equal. This mutant hASIC1a channel regains the same sensitivity to PcTx1 as that of mASIC1a (49). The potentiation of ASIC1b currents by PcTx1 is due to a large shift in the activation curve to less acidic pH (whereas the SSD curve is largely unaffected) and concomitant slowing of desensitisation as seen by an increase in the decay time constant in the presence of PcTx1 (37). The agonist/potentiating effect of PcTx1 on cASIC1 has so far not been studied in much detail but is likely through stabilising the open state of the channel by enhancing activation, slowing desensitisation or both. Interestingly, studies of PcTx1’s functional effect on cASIC1 revealed a biphasic response whereby application of PcTx1 at pH 7.35 transiently opened the channel, followed by only partial desensitisation to a long-lasting, persistent current (38). The potentiating effect of PcTx1 at rASIC1b and cASIC1 may be partly explained by the “pH sensor-trapping model” proposed by Salinas et al. (50). Using ASIC chimeras and mutant channels, it was shown that loss of PcTx1 contacts with either the palm and/or -ball regions of rASIC1a led to current potentiation. This complex pharmacology shows that PcTx1 can preferentially stabilise several states of ASICs that are dependent on the external environment and inherent gating properties of the channel being studied. It also highlights that very small

9 differences in the molecular interaction between peptide and channel can greatly change the functional effect of a ligand.

The substantial subtype, species, and condition-dependent functional effects of PcTx1 are likely to make it a very difficult peptide to manipulate for use as a truly selective pharmacological tool or therapeutic lead, and also warrant caution when interpreting results from in vivo and electrophysiology experiments. Due to its subtype-dependent effects on channel opening and desensitisation, the rate of solution exchange in experimental setups can have a substantial effect on the observed activity and relative potencies. For example, a slow solution exchange of the recording chamber coupled with recording in large oocyte cells (where much of our PcTx1 information has been derived) could lead to inaccurate interpretation of observations if the peptide application is slower than the inherent gating time scale of the ASIC in question. These are less likely to be issues with fast solution exchange as achieved using whole-cell recordings from small cells or macro-patch recordings. Nevertheless, it is important for authors to include such details when reporting their data, and for readers when assessing the conclusions made about an ASIC modulators functional effects.

Using alanine-scanning mutagenesis of PcTx1 the functional pharmacophore at rat (r) ASIC1a was shown to comprise Trp7, Trp24, Phe30, Arg26, Arg27, and Arg28 (see red arrows in Figure 1.2A) (41,51). These findings agree with the co-crystal structures of PcTx1:cASIC1, however the structures showed many more close contacts between the peptide and channel (see blue arrors in Figure 1.2A and the peptide:channel interaction in Figure 1.3A–C) (27,52). From these studies we hypothesise that the peptide binding is largely mediated through the hydrophobic patch comprised of residues Trp7, Trp24, Phe30 (and the C4 sidechain of Arg26) while the functional consequence of binding is due to the positively charged Arg26, Arg27, and Arg28, that form interactions with proton-sensing residues in the acidic pocket (Figure 1.3B) (51), however, this theory requires some direct binding experiments to confirm. An unexpected outcome of these peptide mutagenesis studies was the discovery of some potent ASIC1a potentiating molecules, the PcTx1 mutants R26A and F30A. The subtype selectivity and mechanism of action of these peptides is yet to be determined, however they appear to represent new pharmacological tools with activity profiles different to those so far isolated directly from venoms. The combination of radioligand binding and electrophysiology with chimeric channels (37,53), mutagenesis (51) and co-crystal structures (27,52) has shown that PcTx1 binds to the acidic pocket with the

10 majority of contacts on the thumb domain, with residues on  helix 5 being particularly critical. The thumb region is known to be important for proton binding and subsequent gating of the channel (29), consistent with PcTx1 mimicking the action of protons and inhibiting ASIC1a via stabilising the desensitised state (48).

Figure 1.3: Summary of the active surface of ASIC targeting peptides derived from crystallography and mutagenesis data. (A) Two subunits of cASIC1 are shown as surface representations with the domains colour coded as per Figure 1.1. The partially overlapping binding orientations of PcTx1 (grey) and MitTx (blue and cyan) are shown as peptide backbones. The subunit interface corresponding to the acidic pocket is also highlighted. (B) Surface representation of PcTx1 (grey) bound to cASIC1 showing the extensive hydrophobic interaction with  helix 5 (green residues) and the basic residues (blue) that protrude into the acidic pocket. (C) Surface representation of PcTx1 NMR structure with crystal contact residues highlighted in blue and red. Of these, red represents residues that are the functional pharmacophore and were found to make energetically important interactions for rASIC1a inhibition (these residues are labelled). (D) Active surface of APETx2 for rASIC3 inhibition, mutations of the red residues have a major impact while mutation of the orange had a less severe impact on activity. (E) Active surface of MitTx based on predictions from crystal contacts. (F) Amino acid residues shown to be important for Ma-1 inhibitory activity mapped on to the NMR structure of Ma-2. (NB: structures are not to scale). PDB codes are as follows: cASIC1- PcTx1 complex 4FZ1; MitTx 4NTW; PcTx1 NMR structure 2KNI; APETx2 2MUB; Ma-2 2MFA. 11 PcTx1 and structural insights into ASICs Two groups independently solved the PcTx1-cASIC1 crystal complex in mid 2012 (27,52), and these were the first ligand bound structures of ASICs (Figure 1C). Dawson et al. used the non-functional cASIC1 construct at pH 5.5, which resulted in a channel structure almost identical to the apo structure of this construct (52). This study provided the first detailed picture of the PcTx1 binding site but unfortunately little new information on different channel states. In contrast Baconguis and Gouaux used the 13 cASIC1 functional construct (solved at pH 7.25 and 5.5), and were able to produce crystals in what appear to be two different open conformations. The pH 7.25 structure was proposed to be non-selective for + monovalent cations, whilst the pH 5.5 structure to be Na -selective (PNa/PK = 10/1, i.e. the same selectivity as seen for typical ASIC currents) (27). It remains to be seen if there is any correlation between the two channel states observed independently in the functional studies by Smith and Gonzales (2014) and these structural studies, and what affect the 13 truncation has on channel structure-function (38). Interestingly, the more severe 25 truncation does show less Na+ selectivity and a decrease in the activation curve Hill slope compared to full length channels (28).

At the time, the low pH structure was proposed to be the ‘open’ state of the channel. There is very little difference in the extracellular domain between the ‘open’ and desensitised states, however significant conformational rearrangements occur in the TM domains, as previously predicted from accessibility studies with cysteine-reactive reagents (54). The available crystallographic data suggested desensitisation gating (open to desensitised transition) to involve binding of stimulatory ligands (protons and/or PcTx1) to the thumb and finger, that results in rotational and flexing movement of the TM domains mediated via the wrist and lower palm. How well these structural rearrangements translate to changes that occur during activation gating remains unexplained. Further details on the structure of the pore and ‘open’ state of ASICs were revealed when the MitTx-cASIC1 crystal complex was solved two years later in 2014.

The neuroprotective properties of PcTx1 Being the first-discovered, selective inhibitor of ASICs, PcTx1 has been extensively used in animals to help determine the role of ASIC1a in a wide variety of physiological and pathological processes such as pain, anxiety, depression, epilepsy (with contrasting results) and respiratory control (9,31,32,55). However, the standout condition in which PcTx1 has played a role in validating ASIC1a as a promising therapeutic target is ischaemic stroke. A

12 key role for ASIC1a in the neuronal cell death following ischemic stroke was first reported in 2004 (56). Xiong and colleagues provided evidence from a combination of genetic and pharmacological approaches using the non-selective ASIC inhibitor amiloride as well as PcTx1. However, instead of using pure peptide, “PcTx1 venom” was employed in both the in vitro and in vivo aspects of the study. The crude tarantula venom that contains PcTx1 (at only 0.4%) is complex and contains a multitude of other bioactive peptides (see (57)). “PcTx1 venom” was also used in a follow up study that demonstrated a substantial therapeutic time window for ASIC inhibition (up to five hours post insult) as well as neuroprotective efficacy when delivered intranasally (58). Pure PcTx1 has since been shown to be neuroprotective in both rodent and porcine models of cerebral ischemia when administered 30 min prior to induction of stroke (59). More recently a study confirmed that a single dose of pure PcTx1 (1 ng/kg, i.c.v. 2 hours post stroke) does indeed provide substantial neuronal and functional protection in ischemic stroke in conscious hypertensive rats (57). Promisingly, none of the studies using PcTx1 in vivo reported any obvious adverse effects let alone acute toxicity, consistent with the original CNS toxicity test in mice by Escoubas et al. (34). Thus PcTx1 has proved to be an important tool in the identification and confirmation of ASIC1a as a stroke target.

1.2.2 Hm3a

At the same time as the identification of PcTx1, a second, homologous peptide was identified from tarantula venom (Escoubas unpublished). However, it remained unstudied and unreported until several years later (see Chapter 7), when -TRTX-Hm3a (Hm3a) was isolated from the Togo starburst tarantula (Heteroscodra maculata) (60). Hm3a has five amino acid substitutions compared to PcTx1 and is shorter by three residues at the C- terminus (Figure 1.2A). Of these residues, only the R28K substitution lies within the active site, however is a conservative change that has little apparent effect on potency. To further understand the ASIC1 isoform selectivity and peptide-channel interaction surface for Hm3a and PcTx1, crystal contact residues on the complementary surface to the thumb domain were studied (i.e., the loop between -strands 3 and 4 in the palm domain of the adjacent subunit). Rat ASIC1a was mutated to mimic ASIC1b within the 19-residue region N-terminal to the ASIC1a/1b splice site known to be crucial for PcTx1 function (37,53). Testing of these mutants for sensitivity to PcTx1 identified Arg175 and Glu177 (Cys and Gly in ASIC1b, respectively) to play a key role in the subtype dependent effects of Hm3a. Er et al. also showed that Hm3a potentiates acid-induced currents in oocytes co-injected with rASIC1a

13 and rASIC1b (60), an effect we observed for PcTx1 (unpublished observations), which has not previously been noted.

Interestingly, this study performed in vitro stability assays and revealed that PcTx1 is not as biologically stable as one might assume for an ICK peptide whereas Hm3a is. Hm3a was extremely stable in human serum over 48 hours (~87% remaining), whereas ~35% of PcTx1 remained over the same assay time. Hm3a had very little breakdown over 48 hours at 55 °C, however PcTx1 showed some loss (~24%). The superior stability of Hm3a over PcTx1, coupled with the similar pharmacology of the peptides makes Hm3a a potentially more attractive ASIC tool for in vivo studies.

1.2.3 Hi1a

The recent discovery of a third ASIC inhibitor from spider venom has defined a novel class of ASIC modulatory peptides with unique structural and pharmacological properties. The sequence of Hi1a was noticed during analysis of the venom-gland transcriptome of an Australian funnel-web spider (Hadronyche infensa) as its N-terminus bore a striking similarity to PcTx1 (62% identity). However, unlike PcTx1, Hi1a is almost twice the length at 75 residues, with the C-terminal half also resembling PcTx1 (50% identity; Figure 1.2A) (43). Recombinant production of the peptide allowed structural and functional studies confirming that Hi1a is indeed a highly potent inhibitor of ASIC1a (IC50 of 400–500 pM) and takes the form of two ICK motifs in series (Figure 1.2B), similar to the TRPV1 agonist DkTx (61). Despite Hi1a’s high sequence identity with PcTx1 and Hm3a, it has a unique pharmacological profile against ASICs. In contrast to the single ICK peptides, which affect ASIC1 desensitisation gating (either facilitating it, ASIC1a, or hindering it, ASIC1b) in a pH- dependent and rapidly reversible manner, Hi1a selectively inhibits the activation of ASIC1a in a slowly reversible manner and has little effect on ASIC1b up to 1 M. Thus, Hi1a is the most potent and selective inhibitor of ASIC1a discovered to date making it a valuable research tool. Indeed, the initial assessment of its in vivo efficacy in ischaemic stroke showed that a single dose of Hi1a (2 ng/kg, i.c.v.) provides a very substantial level of neuroprotection even when administered up to eight hours post-insult. Preliminary structure- activity studies show that the binding site of Hi1a overlaps that of PcTx1 and suggest that the relative orientations of the two knot motifs is important for its unique mechanism and kinetics (43). The combined ability of Hi1a to stabilise the resting state of ASIC1a and dissociate slowly suggest that it will be very useful in ASIC structural studies as well as localisation studies of the channel. 14 1.2.4 APETx2

The first selective ASIC3 inhibitor: APETx2 is a 42-residue peptide, isolated from extracts of the sea anemone Anthopleura elegantissima (Figure 1.2A). It was the first potent and selective inhibitor of ASIC3- containing channels with an IC50 of 63 nM for homomeric rASIC3 and 0.1–2 M for heteromeric rASIC3 containing channels (62). The peptide is slightly less potent on the human ASIC3 subtype with an IC50 of 175 nM. APETx2 inhibits the typical transient acid- induced current and the pH 7.0 evoked window (sustained) current of ASIC3 (1,63), as well as the alkali-induced current of hASIC3 (12), but does not inhibit the larger sustained current evoked at pHs lower than 5.0 (62). The structure of APETx2 consists of a compact hydrophobic core composed of a four-stranded β-sheet containing three disulfide bonds, resembling a defensin-like fold (Figure 1.2B) (42,64). At higher concentrations APETx2 also inhibits the voltage-gated sodium channels NaV1.2, NaV1.6, and NaV1.8 with varying degrees of potency depending on subtype and study (65,66). Additional off target effects have been shown with APETx2 also inhibiting the cardiac hERG channel in the low micromolar range (42), demonstrating that this prototypical “selective” ASIC3 inhibitor is not as selective as at first thought.

The pharmacophore of APETx2 for ASIC3 has recently been shown to comprise a contiguous surface made up of loops 2 and 4 and the N-terminus (Figure 1.3D) (42,67). Mutagenesis of APETx2 and its interaction with hERG has also been studied in some detail showing a partial overlap in the pharmacophore with that of ASIC3 (42). Nevertheless, the molecular details of the interaction between APETx2 and ASIC3 (or indeed its off target channels), including the mechanism of action and binding sites have not been extensively studied. It has been speculated that APETx2 binds to the acidic pocket and shifts the channel's apparent affinity for protons (31) in the opposite direction to that of PcTx1. More recently a docking study predicted the binding site to be either the upper thumb area, or slightly above the transmembrane domains, between the wrist and palm region (68). However, there is currently no published functional evidence to support either of these hypotheses. Given that APETx2 is still the most potent ligand available to study ASIC3, unequivocally determining its binding site and mechanism of action are crucial to develop more selective analogues and further our understanding of ASIC3 structure and in vivo function.

15 An APETx2 analogue from another sea anemone -AnmTX Hcr 1b-1 (also Hcr 1b-1) is a peptide related to APETx2 (51% identity, Figure 1.2A) and was isolated from the sea anemone Heteractis crispa (69). The peptide inhibits the transient current of hASIC3 expressed in Xenopus oocytes with an IC50 value of 5.5 M. Although much less potent than APETx2 (~35 fold) and significantly different in sequence, it is interesting to note that the peptides share identity for much of the APETx2 pharmacophore for ASIC3. The sequence differences between these peptides and their comparative activity on APETx2’s off targets should provide some novel insight to structural elements that give rise to the ASIC3 activity profile.

APETx2 in peripheral pain The use of APETx2 in vivo has been key in establishing the role of ASIC3 as a sensor of acid-induced and post-operative pain (63), and demonstrating its involvement in inflammatory pathways (70). In agreement with this, several independent labs have reported that APETx2 is analgesic in rat models of inflammatory and osteoarthritic pain (70-72). Peripheral application of APETx2 also reduces mechanical hypersensitivity in non- inflammatory muscular pain models and angina (73,74). In a study assessing the biological stability of APETx2 and cyclised analogues, the wild-type peptide was almost completely broken down in simulated gastric fluid within 20 minutes while the stability in serum was not assessed (75). Despite the promising results from animal pain models, the biological stability and off target profile of APETx2 would appear to greatly restrict its value as a therapeutic lead candidate.

1.2.5 MitTx

A potent ASIC agonist that causes pain MitTx was isolated from the Texas coral snake (Micrurus tener tener) based on its ability to robustly activate a population of rat cultured sensory neurons (76). The active toxin is a 1:1 complex of two non-covalently bound subunits — MitTx-α consisting of a 60-residue Kunitz type peptide, and MitTx-β, which is a 120-residue phospholipase A2 (PLA2)-like protein (Figure 1.2A). Although inactive as individual components, the heterodimer is a potent pH- independent agonist of ASIC1a and ASIC1b (EC50 9 and 23 nM, respectively), whilst being considerably less potent at ASIC3 (EC50 830 nM). MitTx is a weak outright agonist at ASIC2a (i.e., does not directly activate the channel to more than 10% of pH-evoked current size), however it is a strong potentiator of this channel (amplifying the proton-induced current by ~3 orders of magnitude) by shifting the activation curve to more alkaline pH. 16 The bite of a Texas coral snake is known to be extremely painful, and often requires hospitalisation and administration of opiates (77,78). Peripheral administration of MitTx into mice hindpaws evoked robust nocifensive responses (76). This pain behaviour was also observed in ASIC3-/- mice, but not in ASIC1-/- mice. Furthermore, analysis of the MitTx activity in trigeminal neurons from wild-type, ASIC1-/-, and ASIC3-/- mice revealed the activation was abolished only in ASIC1-/- animals. These experiments provided the first clear evidence of the involvement of ASIC1 in peripheral nociceptive pathways. Previous studies using PcTx1 and various genetic models have yielded inconsistent phenotypes and somewhat inconclusive results in regards to the role of ASIC1 in peripheral nociception (13). Thus, the discovery of the first potent ASIC1 agonist provided a clear leap in knowledge in this field. However, MitTx is a potent agonist of both ASIC1a and ASIC1b channels, making it unclear at this stage which subtype is a more important target for peripheral analgesia.

The true open state structure of ASICs? The 13 cASIC1 construct is sensitive to the agonist properties of MitTx, which lead to solving of the MitTx-cASIC1 co-crystal complex (Figure 1.1D) (26). Significant differences to previous ASIC structures in the TM regions were observed, and this orientation has been interpreted as the open state that most likely resembles the native Na+-conducting state. In this structure there is a break in the α helix of the second TM domain (TM2) that causes it to adopt an extended conformation. This structural rearrangement allows the top half of TM2 from one subunit to interact with the bottom half of TM2 from an adjacent subunit (see Figure 1.1A and D). Interestingly, the TM domain swap was also observed in the 2009 desensitised apo structure upon reanalyses of the electron densities (25), but not in either of the PcTx1 complex structures (27). This swapping of TM domains results in a continuous structure that contains a GAS motif that is important for creating the Na+-selectivity filter of ASICs (54,79- 81). Cs+-soaked MitTx-cASIC1 crystals revealed coordination of Cs+ ions by carbonyl atoms of Gly443 of the GAS motif. The pore dimensions in the selectivity filter of these crystals match that of a hydrated Na+ ion extremely well, suggesting that the TM orientations in this structure may be functionally relevant and provide a barrier mechanism for ion selectivity of ASICs whereby the pore is large enough to admit a hydrated Na+ ion but exclude a hydrated K+ ion. This is inconsistent with the findings of Yang and Palmer who, using hASIC1a, showed that when Na+ is absent, K+ could freely pass through the channel (82). This argues against a barrier mechanism for ion selectivity (or suggests that pore size is not the only factor regulating ion conductance), at least for the human isoform, and highlights that care

17 should be taken when applying observations across specific species or subtype variants of ASICs.

The MitTx-cASIC1 crystal structure not only provided possible further insights into channel gating, but also revealed the expansive MitTx binding site, and both peptide and channel residues that could be functionally critical. In contrast to PcTx1 that binds at the interface between adjacent subunits, MitTx binds almost exclusively to a single subunit from the wrist to the knuckle, along the entire length of the thumb domain. Nevertheless, there is an overlap in the binding site of the  subunit of MitTx and PcTx1 at the acidic pocket (Figure 1.3A), explaining the mutually exclusive functional activity of these peptides observed in earlier studies (76). The  subunit binds below the  subunit towards the base of the thumb and extracellular domains, another region known to be important for channel gating (Figure 1.1D and 1.3A) (83-85).

The various crystal structures of ASICs, channel alone and venom peptide bound, have provided valuable insight into the structural basis of channel gating. Nevertheless, many details regarding the gating of ASICs remain to be determined. All ASIC structures solved to date have been of cASIC1, which can be pharmacologically quite different when compared to mammalian ASIC1a despite ~89% sequence identity. For example, the functional effects of PcTx1 are more similar for ASIC1b and cASIC1 than ASIC1a and cASIC1. And more strikingly, 2-guanidine-4-methylquinazoline (GMQ) has the same effect on the pH activation curve of ASIC3 and cASIC1 (shifting to more alkaline values) but the opposite effect on ASIC1a (38,86). Several of the key questions remaining in this area are: Which ASIC subtype does cASIC1 truly resemble in structure and function, and what does this mean for interpreting data from mammalian channels in light of the cASIC structure? Is there only one true open state, or are there several physiologically relevant open states?

1.2.6 Mambalgins

Pharmacology and structure of mambalgins Mambalgins are a group of homologous peptides isolated from the venom of mamba snakes. Two 57-residue peptides were isolated from the black mamba (Dendroaspis polylepsis polylepsis) that differ by one amino acid have been named mambalgin-1 and mambalgin-2 (Ma-1 and Ma-2, respectively) (18). A third peptide, Ma-3, from the green mamba (Dendroaspis augusticeps) was also isolated and has a single mutation at position 23 when compared to Ma-1 (Figure 1.2A) (31). These peptides conform to the three-finger

18 toxin fold (Figure 1.2B) and inhibit homomeric ASIC1a (rat and human IC50 3–55 and 127 nM, respectively) and rASIC1b (IC50 44–192 nM), as well as ASIC1a- and ASIC1b- containing heteromers with weaker potency (18,44,45). Mambalgins inhibit ~60% of ASIC currents in rat sensory neurons, whereas PcTx1 only inhibits ~40%, most likely due to the broader activity profile of mambalgins through ASIC1b containing channels (18). A study using human stem cell-derived sensory neurons showed that Ma-1 also inhibits native human ASIC currents (87). Although there are small discrepancies in the potency of peptides across platforms and publications, all three peptides have been reported to have the same overall pharmacological profile.

Mambalgins are thought to preferentially bind to the closed state of ASICs and inhibit channels by shifting the pH-dependence of activation to more acidic pH, decreasing their apparent affinity for protons (50,88). Chimeric studies in which domains of ASIC1a were swapped with ASIC2a (not sensitive to mambalgins) suggest that the mambalgins bind in the region of the acidic pocket, with the peptide interacting with the -ball and upper thumb of one subunit and the palm of the adjacent subunit (50). Single point mutants of the channel in this region were also made to identify specific interacting residues of which Phe350 on helix 5 in the thumb stands out as an important contact (45,50). However, these studies lack the detail to explain the observed selectivity/potency differences between ASIC species isoform and subtypes. The pharmacophore of Ma-1 has also been partially determined with alanine mutants only made so far in loop 2. Mutating residues Phe27, Arg28, Leu32, Ile33, and Leu34 induced a significant decrease in potency, and this face of the peptide was determined to be the ASIC-interacting surface (Figure 1.3F) (44), in contrast to the results of a prior docking-based prediction by the same group (50). A third study which solved the cryo-EM structure of the Ma-1-cASIC1a complex at 5.4 Å suggest mambalgins bind predominantly to the thumb domain of ASICs (89). Due to the low local resolution of the cryo-EM structure, de novo structure determination was not possible and the authors rigid body fit the density of the desensitised cASIC1 (PDB code 4FZ1) and Ma-1 (PDB code 5DU1) crystal structures into the cryo-EM density map. Overall these studies propose vastly different peptide binding poses and provide insufficient detail to explain the ASIC species and subtype selectivity of mambalgins.

Unlike PcTx1 and APETx2, recombinant production of mambalgins has not been reported. However several elegant strategies have been developed to synthesise the peptide for structural and functional studies (44,45,90). Two high-resolution NMR structures have been

19 published and are in good agreement with one another, with the lower part of loop 2 showing some flexibility within the NMR ensemble (45,90). More recently crystal structures of two Ma-1 polymorphs were determined, showing significant structural variations compared to the NMR structures (see Figure 1.2B) (44). In both crystal structures the lower part of loop 2 is extended and well defined and there was a large degree of conformational variability in loop 3. Of the two crystal structures, the most frequently observed and most ordered conformation of loop 3 was a single-turn helix. The NMR structures of this region contain a short β-sheet, and this conformation can also be seen as a less frequent form within the crystal structures. It must also be noted that for both polymorphs of the wild-type peptide, the Ma-1 molecules assembled together as complexes of either dimers or tetramers in a single asymmetric unit. In both crystals there were significant inter-molecular interactions present, which may explain the differences to the more flexible NMR structures, and this may result in structures where the Ma-1 loops have been stabilised in non-physiologically relevant conformations. These differences between the NMR and crystal structures are extremely important when studying the functional surface of the peptide, as they yield different surface-exposed residues. This will have profound implications when mapping on pharmacophore residues and understanding the nature of the interaction of these peptides with ASICs. Unfortunately, no co-crystal structure of mambalgins in complex with ASICs has been reported to date. Either structural information of the mambalgin-ASIC interaction, or more extensive peptide and channel mutagenesis will be required to truly understand this interaction to facilitate design of ASIC1a or ASIC1b selective analogues.

Is ASIC1b or ASIC1a important in peripheral nociception? The discovery that MitTx robustly elicits nocifensive (pain-related) behaviour in mice upon intraplantar injection highlighted ASIC1 as an important mediator of peripheral nociception. However, the lack of selective inhibitors of ASIC1b, and lack of peripheral analgesic activity of PcTx1 (91) made it difficult to tease out the relative importance of ASIC1a and ASIC1b in peripheral pain. Ma-1 was found to be analgesic when administered centrally or peripherally and experiments with ASIC knockout and knockdown animals suggest that the central effects of Ma-1 are mediated by ASIC1a, whereas peripheral analgesia is mediated by ASIC1b (18). This group has now demonstrated the analgesic potential of mambalgins in rodent models of thermal, inflammatory and neuropathic pain, however this finding has yet to be reported by any independent group (18,88).

20 1.2.7 Ugr 9-1

Sea anemones have proved to be a consistent source of ASIC ligands with the painted anemone (Urticina grebelnyi) yielding Ugr 9-1 (also -AnmTx Ugr 9a-1) (Figure 1.2A), a low affinity inhibitor of both transient and sustained current of hASIC3 (46). The peptide reversibly and completely blocked the transient current with an IC50 of 10 M, and inhibited the sustained current with maximal inhibition of ~48% and an IC50 of ~1.4 M. No activity was observed at up to 50 M on ASIC1a, ASIC1b, and ASIC2a, potentially making it a more selective ASIC3 inhibitor than APETx2. In this study the peptide was applied for only 10 s before the low pH stimulus, thus, depending on its on-rate the IC50 values determined could be underestimating the true potency (i.e. if steady-state inhibition is not reached within the short application time used). The peptide was successfully produced in E. coli to assist in structure determination and animal studies with a final yield of ~8 mg of peptide per litre of culture. The NMR structure revealed a novel scaffold for short peptides named the boundless -hairpin. The Ugr 9-1 structure is ‘boundless’ as the disulfide bonds do not stabilise the core of the peptide (no interstrand disulfides), instead one disulfide links the N- terminus to the -core, and another the C-terminus to a -turn (Figure 1.2B).

Assessment of the Ugr 9-1 three dimensional surface, and sequence comparison with the ASIC3 inactive homologs Ugr 9-2 and Ugr 9-3 led Osmakov et al. to predict the functional surface to contain Phe9, His12, and Tyr24 of Ugr 9-1 (as these are the only three residues that are unique to Ugr 9-1 when compared to Ugr 9-2 and Ugr 9-3) (46). Follow up work from the same laboratory introduced the individual mutations T9F and Y12H into the inactive Ugr 9-2 to resemble the active Ugr 9-1 (92). The mutant peptides gained ASIC3 activity but were 2.2- and 1.3-fold weaker than native Ugr 9-1, respectively, demonstrating the importance of these residues for inhibitory activity. An N- and C-terminally truncated analogue of Ugr 9-1, Ugr22, was also produced and shown to be equipotent with native Ugr 9-1 at inhibiting ASIC3. The mechanism of action and binding site of these peptides is still unknown.

Ugr 9-1 showed analgesic effects in two models of inflammatory pain, the CFA-induced thermal hyperalgesia model, and the acetic acid writhing test (46). Unsurprisingly, neither of the ASIC inactive peptides Ugr 9-2 and Ugr 9-3 had any effect in the hyperalgesia model. Animals treated with any of the three active peptides showed no motor impairment or sedation, making Ugr 9-1 and Ugr22 promising tools to study ASIC3 in vivo. Despite their

21 comparable potency at ASIC3, Ugr22 reversed thermal hyperalgesia with greater efficacy than Ugr 9-1 at 0.1 mg/kg dose (92) possibly due to the smaller mass of Ugr22, therefore higher molar dose.

1.2.8 PhcrTx1

PhcrTx1 (also -PMTX-Pcf1a) was isolated from the sea anemone Phymanthus crucifer as an inhibitor of ASIC currents in rat DRG neurons with an IC50 ~100 nM (93). It also partially (less than 20%) inhibits voltage-gated K+ currents in DRG neurons in the low micromolar range (IC50 ~3.5 M). PhcrTx1 is a partial inhibitor of the native ASICs currents with maximal inhibition of ~44% at 10 M. PhcrTx1 is fully reversible within 5 minutes, did not affect the current desensitization rate, and was not active when applied only with the stimulus pH. This suggests the peptide preferentially binds to the closed state of ASICs and prevents activation, potentially providing another novel tool to help understand channel gating. Given the relative potency in inhibiting ASIC currents in rat DRGs, it would be of interest to determine the subtype selectivity of PhcrTx1 and its analgesic potential. PhcrTx1 appears to be the first ICK peptide isolated from sea anemones and only shares 32% identity with PcTx1 (Figure 1.2A), however the structure or disulfide bond connectivity have not yet been determined experimentally. Nevertheless, molecular modeling of PhcrTx1, calculation of the dipole moment, and comparison with the PcTx1 pharmacophore led the Rodríguez et al. to predict a cluster of basic/aromatic residues to be the functional site for ASIC activity (93).

1.2.9 -Dendrotoxin

-Dendrotoxin (-DTx) is another toxin isolated from the green mamba venom (same source as Ma-3), and is a well-known low nanomolar KV1.x channel blocker that was recently shown to also inhibit ASICs with weaker potency (94). The 59 amino acid peptide is a single Kunitz domain fold (47), similar to the -subunit of MitTx (Figure 1.2A). They only share ~32% identity and 55% similarity, however four of the six identified crystal contacts for MitTx are similar in -DTx raising the possibility that they may bind to a similar site on ASICs (albeit with different functional outcomes). -DTx reversibly inhibited ASIC peak currents in rat

DRG neurons with an IC50 of 800 nM. Although the binding site and mechanism of action have yet to be determined, -DTx inhibition was only present when cells were preconditioned with peptide (applied for 20 s during conditioning at pH 7.4, and also for 5 s during the pH 6.1 acid pulse), but not when applied only with the stimulating pH. This may

22 suggest that -DTx preferentially binds to the closed state of the channel and prevents activation.

1.2.10 Conorfamide RPRFa (also CNF-Tx1.1)

The first ASIC modulators from cone snails were identified when three short RFamides that potentiate and slow desensitisation of ASIC3 currents were isolated from the venom of Conus textile (Figure 1.2A) (95). The most potent was the shortest analogue with just four amino acids, RPRFa (Arg-Pro-Arg-Phe-NH2), with an EC50 of 4.23 M for potentiating the

ASIC3 late current (I5s/Ipeak). RPRFa was also shown to increase DRG excitability and acid- induced muscle pain in mice. Subsequent work showed that RPRFa binds to the closed state of ASIC3, remains bound in the open state to slow desensitisation of proton-evoked currents, and its unbinding allows for complete desensitisation of ASIC3 (96). Mutations to residues of the nonproton ligand-sensing domain within the ASIC central vestibule were shown to reduce the activity of RPRFa, indicating peptide binding to this cavity prevents collapse of this region that is normally associated with desensitisation gating. RPRFa was crucial in acting as a positive modulator of ASIC3 and confirming the importance of this ASIC subtype in peripheral pain sensing. The novel pharmacological profile in comparison to other ASIC ligands will almost certainly provide value as a research tool for future studies of ASIC3.

1.3 CONCLUSIONS

Since the discovery of PcTx1 at the turn of the millennium, our understanding of ASIC modulating venom peptides themselves, and their use in deciphering ASIC structure- function and involvement pathological processes have greatly increased. It is now quite clear that as for many other ion channel families, animal venoms are a rich source of potent and valuable ASIC modulators. As detailed above, there are now eleven known ASIC modulators from venoms (not including close homologs), so far from only four sources, spiders, sea anemones (the most prolific source to date), snakes, and cone snails. Despite this relatively low number (when compared to the number of venom peptides known to modulate NaV or

KV channels), a trend is starting to appear. The modulators from snakes and spiders are, so far, relatively selective for ASIC1, while the sea anemone peptides seem to preferentially target ASIC3. Whether this holds true over the years to come and what it means from an ecological point of view remain to be seen. Whether venoms will harbour any components that have selectivity for ASIC2 (which to date suffers a dearth of selective tools) also remains

23 to be seen. In the mean time we still have a lot to learn about the tools described above in order to maximise their value as research tools.

Several small molecule modulators of ASICs characterised have poor selectivity against ASIC subtypes. Given the sometimes extensive interaction surface and multi-domain nature of venom peptide binding, we believe these gifts from nature are our best chance at discovering and developing more selective pharmacological tools that can better discriminate between ASIC subtypes. Gaining a thorough understanding of the molecular interactions between selective venom peptides and ASICs will almost certainly facilitate answering some of the major unresolved questions in the ASIC field and may even lead to some novel therapeutics along the way.

1.4 THESIS AIMS

The overall goal of this thesis was to provide a better insights into how several venom peptides interact with ASICs, in particular their mechanisms of action and molecular basis underlying selectivity. The specific aims are as follows:

1. Determine the reason for the difference in PcTx1 potency at human and rat ASIC1a. Furthermore to probe the amino acid interactions that differ between PcTx1 and rASIC1a, rASIC1b, and hASIC1a variants.

2. Investigate the mechanism of action and binding site of the ASIC3 inhibitor APETx2.

3. Delineate the molecular basis for potency and selectivity of mambalgins at species and subtype variants of ASIC1, and clarify the disagreement in published models for the mambalgin binding site at ASIC1a.

4. Discover and characterise novel venom peptide modulators of ASICs.

Together, the aims of this thesis provide information on important ASIC research tools, namely venom peptide modulators, and expand the knowledge in the field of ASIC pharmacology so they can be better used as research tools.

24

CHAPTER 2

Materials and Methods

25 2.1 RECOMBINANT PEPTIDE PRODUCTION

Animal venoms are a rich source of secreted disulfide-rich peptide (DRP) (97,98). The major challenge for production of recombinant DRPs is obtaining the native disulfide-bond isomer in reasonable yield. A DRP containing three disulfide bonds (six cysteines) can theoretically form 15 different disulfide-bond isomers, and the number of possible isomers rapidly increases to 105 for four disulfide bonds. Several peptides studied in this thesis (Table 2.1 for summary of wild type peptides only) were produced using the periplasmic Escherichia coli expression system outlined below. The major advantage of this system is that the periplasm houses the molecular machinery for disulfide-bond formation (Figure 2.1) (99- 101).

Table 2.1: Disulfide-rich venom peptides expressed for this thesis. S-S Yield Toxin name Organism Residues Target Ref. bonds (mg/L) π-TRTX-Pc1a Spider 40 3 >5 ASIC1 (41) Sea APETx2 42 3 0.5–1.0 ASIC3 (67) anemone Mambalgins Snake 57 4 0.1 ASIC1 N/A π-TRTX-Hm3a Spider 37 3 1 ASIC1 (60)

Figure 2.1: Schematic of the pathway for production of secreted proteins in E. coli. After translation of the encoding mRNA in the cytoplasm, the protein is translocated into the periplasm. During translocation, the signal sequence is removed to release the mature protein. The periplasmic Dsb system, comprised of the DsbA, DsbB, DsbC and DsbD proteins, subsequently aids in disulfide- bond formation. Adapted from Ref. (102).

26 2.1.1 Plasmid design

A synthetic gene encoding each specific DRP, with codons optimised for E. coli expression, was cloned into a variant of the pLic-MBP expression vector (Figure 2.2A). This IPTG- inducible plasmid produces a MalESS-His6-MBP-DRP fusion protein with a tobacco etch virus (TEV) protease cleavage site directly preceding the toxin-coding region. The MalE signal sequence (MalESS) directs the fusion protein to the periplasm, maltose binding protein

(MBP) enhances toxin solubility, and the His6-tag enables facile purification of the fusion protein via nickel affinity chromatography. The TEV recognition sites leaves a residual N- terminal residue after cleavage. Although several residues have been reported to be tolerated at the P1’ site (103), glycine and serine have most successfully been used (102,104), thus were used for all peptides produced here.

2.1.2 Expression and purification

Plasmids were transformed into E. coli strain BL21(DE3) for recombinant DRP production. Cultures were grown in Luria-Bertani (LB) medium containing 100 g/ml ampicillin at 37°C with shaking at 200 rpm. After cooling cultures to 18°C DRP expression was induced with 0.5 mM IPTG at an OD600 of ~1.0 (Figure 2.2B), and cells grown for 12 h before harvesting by centrifugation (10 min at 5,000 g). The fusion protein was extracted from the bacterial periplasm by cell disruption at 32 kPa (TS Series Cell Disruptor, Constant Systems Ltd, UK), then captured by passing the soluble extract (in 30 mM Tris, 300 mM NaCl, pH 8.0) over Ni-NTA Superflow resin (Qiagen). Nonspecific protein binders were removed by washing with 15 mM imidazole then the fusion protein eluted with 300 mM imidazole. The buffer was exchanged to remove imidazole and the fusion protein concentrated to 10 mL using a 30K spin column. Reduced and oxidised glutathione were added to 3 and 0.3 mM (10, and 1 mM for mambalgins), respectively, to maintain TEV protease activity. Approximately 40 g of TEV protease was added per mg of DRP, and the mixture incubated at room temperature for 12 h (Figure 2.2C).

27

Figure 2.2: Expression of DRPs. (A) The vector coding region includes a MalE signal sequence

(MalEss) for periplasmic export, a His6 affinity tag, a solubility-enhancing maltose binding protein (MBP) fusion tag, and the coding sequence of the selected disulfide-rich peptide or protein (DRP), with a TEV protease recognition site inserted between the fusion tag and the peptide coding regions. The location of key elements of the vector are shown, including the ribosome binding site (RBS) and key restriction sites. (B) SDS-PAGE gel showing espression of fusion protein after IPTG induction. Lanes: 1 and 3, cell extract before IPTG induction; 2 and 4, extract from IPTG-induced cells. (C) SDS-PAGE gel showing key stages during the purification. Lane: 1, soluble extract after cell disruption; 2, fusion protein eluted from Ni-NTA column before TEV protease cleavage (sample is

His6-MBP-DRP); 3, post-cleavage sample (His6-MBP with the cleaved DRP free).

2.1.3 HPLC purification

Cleaved MBP and TEV were precipitated by addition of 0.5% trifuoroacetic acid (TFA), then the solution was centrifuged at 20,000 g for 15 min to remove precipitated material before purification of DRPs using reversed-phase (RP) HPLC. Initial RP-HPLC was performed on a Phenomenex Jupiter C4 column (250  100 mm) using a flow rate of 3 mL/min and a gradient of 10–50% solvent B (0.043% TFA in 90% acetonitrile) in solvent A (0.05% TFA in water) over 40 min. Peptides were then subject to a second round of purification using a Zorbax C18 SP column (250 x 10 mm) using a flow rate of 3 mL/min and a 1% per minute gradient.

28 2.1.4 MALDI-TOF mass spectrometry

Peptide mass was determined using matrix-assisted laser desorption/ionisation time-of- flight (MALDI-TOF) MS using a Model 4700 Proteomics Analyzer (Applied Biosystems, Foster City, CA, USA) collected in both positive reflector and linear mode. All masses reported are M+H+. Fractions were mixed (1:1, v/v) with α-cyano-4-hydroxycinnamic acid matrix (7 mg/mL in 50% acetonitrile in H2O). To obtain a sequence tag of the peptide the reductive matrix 1,5-diaminonaphthalene (1,5-DAN) was used (105). The active fractions were mixed (1:1, v/v) with 1,5-DAN (10 mg/ml in 50% acetonitrile, 0.1% formic acid in H2O). All spectra were acquired in reflector mode and 20 spectra of 50 laser shots were accumulated based on defined acceptance parameters and adequate signal intensity in the 800–5000 m/z range.

2.2 OOCYTE ELECTROPHYSIOLOGY

2.2.1 Two-electrode voltage clamp recordings

X. laevis (Xenopus Express) and X. borealis (Nasco) oocytes were surgically removed from anaesthetized (Tricaine methanesulfonate, MS-222) female frogs and treated with collagenase (1 mg/ml; Sigma type I) for defolliculation. cRNAs were synthesised using an mMessage mMachine cRNA transcription kit (Ambion Inc., Austin, TX, USA) and injected into stage V–VI oocytes at 4–200 ng per cell. Oocytes were stored at 17°C in ND96 solution

(96 mM NaCl, 1.8 mM CaCl2, 2 mM KCl, 2 mM MgCl2, 5 mM HEPES, pH 7.4) supplemented with 2.5 mM pyruvic acid, 50 g/mL gentamicin, and either 2.5% horse serum or 0.5 mM theophylline. Membrane currents were recorded 2–10 days after injection under voltage- clamp at -60 mV unless studying voltage dependent effects (Axoclamp 900A, Molecular Devices, CA, USA) using two standard glass microelectrodes of 0.5–2 MΩ resistance when filled with 3 M KCl solution (digitised 5 kHz and filtered at 0.01 Hz). Data acquisition and analysis were performed using pCLAMP software (Version 9.2 or 10, Molecular Devices, Sunnyvale, CA, USA). Changes in extracellular solutions were induced using a microperfusion system that allowed local, rapid exchange of solutions (bath volume = ~30 L). All experiments were performed at room temperature (18–21°C) in ND96 solution. Experiments containing peptides were performed in ND96 solution containing 0.1% fatty acid free-bovine serum albumin (BSA).

29 ASIC currents were elicited by a drop in pH from 7.45 to 4.5–6.5 (for 5–10 s) every 60 s depending on the channel studied. Unless stated otherwise, peptides were applied at pH 7.45 and oocytes bathed in the peptide containing solution for 50 s before the next low pH stimulation. In most cases currents were normalised to control currents activated (by pH 5 or 6) from a conditioning pH of 7.45 in peptide-free solution. pH-activation curves were generated by using a conditioning pH of 7.45 or 7.9, and stimulating every 60 s with an increasing pH drops from 7.2 to 4.5. Steady-state desensitisation curves were generated by applying different conditioning pH solutions (between pH 6.0 and 7.75) for 120 s prior to stimulation by pH drop to pH 5 or 6.

All data were analysed using Prism 6.0 and 7.0 (GraphPad Software, San Diego, CA, USA). The Hill equation was fitted to normalised concentration-response curves to obtain the half- maximal response (EC50, IC50, or pH50) and Hill coefficient (nH). Voltage-dependent activation curves were fit with the Boltzmann equation to obtain the V0.5 (the voltage corresponding to half-maximal effect). The number of replicates (n) represent separate experimental oocytes taken from more than 2 frogs for each data set. Chapter specific data analyses techniques and protocols are presented as they appear under the Data Analyses sub-section.

2.2.2 Ethics statement

This study was carried out in strict accordance with the recommendations in the Australian code of practice for the care and use of animals for scientific purposes, 8th Ed. 2013. The protocol was approved by the Anatomical Biosciences group of the Animal Ethics Committee at The University of Queensland (Approval Numbers: QBI/059/13/ARC/NHMRC, see Appendix I; and QBI/AIBN/087/16/NHMRC/ARC, see Appendix II). Frog recovery surgery was performed under anaesthesia (animals bathed in 1.3 mg/mL of MS-222), and all efforts were made to minimise suffering. The minimum time between surgeries on the same animal was three months, and on the final surgery (maximum of six) frogs were euthanised by decapitation under MS-222 and ice anaesthesia.

2.3 MOLECULAR BIOLOGY

2.3.1 Mutagenesis

Point mutations were produced using site-directed mutagenesis by PCR using Platinum Pfx polymerase (Thermo Fischer Scientific, Scoresby, VIC, Australia) (106), and chimeras were

30 generated by sequential PCR with Q5 polymerase (Genesearch, Gold Coast, QLD, Australia) (107). All clones were Sanger sequenced to confirm the desired changes and integrity of constructs (Australian Genome Research Facility, Brisbane, QLD, Australia).

31

CHAPTER 3

Xenopus borealis as an alternative source of oocytes for biophysical and pharmacological studies of neuronal ion channels

32 Preface to Chapter 3. This chapter demonstrates that Xenopus borealis oocytes are a viable heterologous system for expression of ion channels, and has been modified from a published manuscript. The GlyR and GABAAR TEVC and VCF recordings has been removed from the submitted manuscript as this data was collected by Ming S Soh and Sahil Talwar in Joseph W Lynch’s lab. I performed all oocyte experiments and data analyses presented in this chapter. I performed the sample preparation for the imaging studies and images were acquired by John DO Griffin and Darren L Brown. I produced recombinant PcTx1 and APETx2, and Zoltan Dekan produced synthetic ProTx-I. The text below taken from the manuscript was written by me and edited by Lachlan D Rash, and all figures prepared by me.

3.1 INTRODUCTION

Xenopus frogs have been extensively used for almost a century as model organisms for a wide range of biological applications. Xenopus have been most extensively used for developmental and genetic studies, as external development of the embryos and their large size allows high throughput studies on a vertebrate system. Gurdon and colleagues were the first to show that injection of heterologous mRNA into Xenopus laevis oocytes resulted in robust expression of exogenous proteins (108). Furthermore, Xenopus oocytes possess much of the cellular machinery required to produce functionally important post-translational modifications. These properties were exploited to show that Xenopus oocytes could faithfully produce surface expressed, functional nicotinic acetylcholine receptors following injection of mRNA from Torpedo marmorata (109). Xenopus oocytes subsequently became a valuable tool for the study of mammalian membrane proteins, receptors, and ion channels particularly in the fields of neuroscience and pharmacology (110,111). With the growing number of genomic studies identifying mutations associated with human diseases, the oocyte system continues to be of importance in determining the functional effect of these mutations in expressed proteins (112,113), and as a tool to screen for and characterise novel modulators of ion channels or receptors known to be involved in poorly treated conditions such as pain (18,42).

There are more than 20 species of Xenopus, but the African clawed frog X. laevis appears to be the only species that has been used for electrophysiological studies of ion channels. X. tropicalis and the Marsabit clawed frog X. borealis are widely used in many laboratories for genetic and developmental studies (114-119). Although X. tropicalis is more widely used,

33 X. borealis is more closely related to X. laevis (120). However, heterologous ion channel expression has yet to be performed using X. borealis oocytes.

Here we present data showing the first use of X. borealis oocytes for electrophysiology studies, and provide a comparison with X. laevis oocytes. We analysed the expression profile of a variety of voltage- and ligand-gated ion channels using two-electrode voltage- clamp and voltage-clamp fluorometry (VCF) methods. Furthermore, we investigated the vitelline and plasma membrane of X. laevis and X. borealis oocytes using scanning electron microscopy (SEM), and observed a clear difference in membrane structure. We found that X. borealis oocytes can be successfully used as a heterologous expression system for neuronal ion channel studies.

3.2 METHODS

3.2.1 Peptide preparation

Recombinant APETx2 and Pc1a were produced as described in Chapter 2 and ProTx-I was synthesised using Fmoc chemistry as described (121).

3.2.2 Data analyses

Data are shown as mean and 95% confidence intervals. Most equivalence testing comparing X. laevis and X. borealis oocytes was performed using an unpaired Student’s t-test with 95% confidence intervals. A Welch’s t-test was performed when two populations were not assumed to have equal standard deviations (e.g. Figure 3.5). For both t-tests, it was assumed data was sampled from Gaussian populations, and P values were calculated from two-tailed tests. P < 0.05 was considered statistically significant, unless stated otherwise.

3.2.3 Scanning electron microscopy

Oocytes were fixed in 2.5% glutaraldehyde in Modified Barth’s Solution (88 mM NaCl, 1 mM

CaCl2, 1 mM KCl, 1 mM MgSO4, 2.5 mM NaHCO3, 5 mM HEPES, pH 7.6) at 4°C for 12 h. Oocytes were then post fixed in 1% osmium tetroxide for 1 h, dehydrated through ethanols and finally dried using hexamethyldisilazane (HMDS) (all from ProSciTech, Queensland, Australia), before being coated with gold. Images were captured using a JEOL JCM5000 Neoscope scanning electron microscope (Laboratory Scientific Engineering).

34 3.3 RESULTS

3.3.1 Comparison of X. laevis and X. borealis oocytes

X. laevis and X. borealis oocytes can be easily distinguished visually by their size and the pigmentation of their vegetal hemispheres (Figure 3.1). Stage V–VI oocytes from X. laevis (1.0–1.3 mm diameter) have a pale yellow (sometimes pale green) vegetal hemisphere. X. borealis oocytes are smaller (0.8–1.1 mm diameter), have a darker animal hemisphere and brown vegetal hemisphere clearly separated by a light-yellow belt around the middle. The resting membrane potential of healthy oocytes expressing ion channels of interest from both species was approximately –35 mV over the period in which electrophysiology recordings were performed (with the exception of voltage-gated potassium (KV) channels that had a more negative resting membrane potential around –60 mV). The microinjection process also revealed that the defolliculated membranes of X. borealis oocytes were stronger and more resistant to injection than those of X. laevis, suggesting the presence of a stronger vitelline membrane (see below).

1.0 mm Figure 3.1: Comparison of defolliculated stage V–VI oocytes from X. laevis (left) and X. borealis (right). Oocytes were imaged on an Olympus SZX12 stereomicroscope in ND96 solution.

One of the main advantages of the X. laevis expression system for ion channel studies is the low level of endogenously expressed channels and receptors, which yields a relatively electrophysiologically silent background. Nevertheless, the most prominent endogenous current in these cells is from the well-characterised Ca2+-activated chloride channel (Figure 3.2A) (122,123). This outward current was also observed in X. borealis oocytes when subject to high voltage pulses (greater than +80 mV) (Figure 3.2B). As has been observed 35 for X. laevis oocytes, the amplitude of this current in X. borealis was batch dependent and present in ~85% of all oocytes tested. However, the average amplitude of this current was significantly smaller in X. borealis oocytes (Figure 3.2C). No other protocols (neither voltage- nor ligand-gated) used in this study to activate the ion channels tested led to the generation of observable currents in naïve oocytes from either X. laevis or X. borealis.

Figure 3.2: Endogenous Ca2+-activated chloride currents in naïve X. laevis () and X. borealis () oocytes. Representative families of whole-cell currents elicited by steps from –120 mV to +100 mV in 20-mV increments in oocytes from (A) X. laevis and (B) X. borealis (holding potential of –100 mV). (C) Peak amplitude as measured at the end of a 2 s pulse to +100 mV. A significant difference was observed between the two species (unpaired t-test, P = 8.3 × 10–5). Each data point indicates recording from a single oocyte (n = 22). Several oocytes were used from each of five individual frogs of each species. Data are presented as mean (dashed line) and 95% confidence intervals.

3.3.2 Voltage-gated ion channels

Two-electrode voltage-clamp electrophysiology recordings were performed to compare the properties of a panel of KV and voltage-gated sodium (NaV) channels heterologously expressed in X. laevis and X. borealis oocytes. The voltage sensitivity of KV10.1 (hEAG) activation (Figure 3.3A) and the threshold of activation in KV11.1 (hERG) did not differ between species (Figure 3.3B). Conductance-voltage relationship curves obtained for

NaV1.2 (Figure 3.3C), NaV1.5 (Figure 3.3D), and NaV1.7 (Figure 3.3E) overlap well between

X. laevis and X. borealis oocytes. Pharmacological modulation of NaV channels was assessed by comparing the inhibition induced by 300 nM ProTx-I, a potent and non-selective

NaV channel inhibitor (124). Comparable levels of inhibition between species were detected for all NaV channel subtypes tested (Figure 3.3F). There is no evidence of a difference in current amplitudes for voltage-gated ion channels expressed in X. laevis as compared to X. borealis (Table 3.1).

36

Figure 3.3: Voltage-dependent properties of KV and NaV channels expressed in X. laevis () and X. borealis () oocytes. (A) Current-voltage relationship showing activation of KV10.1 channels (n =

12). (B) KV11.1 tail current plotted as a function of voltage to show activation properties (n = 9–12).

The normalised deduced conductance (G)-voltage relationships for (C) NaV1.2, (D) NaV1.5, and (E)

NaV1.7 (n = 12–17). (F) The effect of 300 nM ProTx-I on current evoked by a depolarisation to –15 mV from a holding potential of –80 mV (n = 8–9). There is no evidence of a difference between the two species (unpaired t-test, P > 0.05). Error bars indicate 95% confidence intervals.

Table 3.1: Maximal current amplitudes recorded from exogenous channels expressed in X. laevis and X. borealis oocytes. Maximal current ± s.d. P values were calculated comparing X. laevis and X. borealis using an unpaired t-test. Ion Channel Maximal current (A) P value n number X. laevis X. borealis

KV10.1 21.0 ± 11.8 23.6 ± 8.5 0.49 15

KV11.1 1.1 ± 0.8 0.9 ± 0.2 0.31 14

NaV1.2 5.3 ± 3.6 8.0 ± 5.3 0.16 11

NaV1.5 3.3 ± 3.0 5.1 ± 2.5 0.08 15

NaV1.7 2.2 ± 1.4 1.9 ± 0.5 0.38 12 ASIC1a 4.8 ± 4.5 3.9 ± 3.0 0.44 20 ASIC1b 3.5 ± 2.9 5.5 ± 4.2 0.13 20 ASIC2a 3.8 ± 3.0 3.8 ± 3.7 0.96 18 ASIC3 1.2 ± 1.1 1.3 ± 0.8 0.94 23 37 3.3.3 Ligand-gated ion channels

The two-electrode voltage-clamp method was also applied to a range of ligand-gated ion channels expressed in X. laevis and X. borealis oocytes. The pH-sensitivity of activation and steady-state desensitisation for homomeric proton-sensitive acid-sensing ion channels (ASICs) were similar between species for ASIC1a (Figure 3.4A), ASIC1b (Figure 3.4B), ASIC2a (Figure 3.4C), and ASIC3 (Figure 3.4D). To assess any differences in the pharmacological properties of ASICs expressed in oocytes from each species, the activity of two known peptide inhibitors of ASICs was tested. -TRTX-Pc1a (Pc1a, also known as

PcTx1) inhibited rat ASIC1a expressed in both X. laevis and X. borealis oocytes with an IC50 of 1 nM (Figure 3.4E), which corresponds well with the reported literature value of 0.9 nM obtained using X. laevis oocytes (34). APETx2 inhibited rat ASIC3 expressed in X. laevis and X. borealis oocytes with an IC50 of 65 nM and 52 nM (Figure 3.4F), in good agreement with the reported value of 61 nM for APETx2 in X. laevis oocytes (62). There is no evidence of a difference in current amplitudes for ligand-gated ion channels expressed in X. laevis as compared to X. borealis (Table 3.1).

Figure 3.4: The effect of activating and antagonist ligands on different ASIC subtypes expressed in X. laevis () and X. borealis () oocytes. pH-dependence of steady-state desensitisation (open symbols, dashed lines) and activation (closed symbols, solid lines) of (A) ASIC1a, (B) ASIC1b, (C) ASIC2a, and (D) ASIC3 (n = 9–12). Concentration-effect curve for inhibition of (E) rat ASIC1a by Pc1a (n = 9–12) and (F) rat ASIC3 by APETx2 (n = 9). Error bars indicate 95% confidence intervals.

38 3.3.4 Scanning electron microscopy

It was apparent that X. laevis and X. borealis oocytes have distinct differences in their vitelline membrane properties. To gain further insights into this difference, the oocyte surfaces were imaged using scanning electron microscopy (SEM) at each time point during the protease treatment procedure. X. laevis oocytes without protease treatment (Figure 3.5A and D) and after 1 min treatment (Figure 3.5B and E) were very similar. In contrast, treatment for 1.5 min (Figure 3.5C and F) completely removed the vitelline membrane and exposed the underlying plasma membrane. At low magnification, untreated X. borealis oocytes (Figure 3.5G) largely resemble untreated X. laevis oocytes. However, protease treatment for 1 min results in a clearly wrinkled vitelline membrane (Figure 3.5H and K) and after 1.5 min of treatment the plasma membrane is exposed (Figure 3.5I and L).

39

Figure 3.5: SEM images of defolliculated X. laevis and X. borealis stage V–VI oocytes prior to and following protease treatment. Oocytes from X. laevis (A-F) and X. borealis (G-L) were viewed by SEM following treatment with protease for 1min (B,E,H,K) and 1.5mins (C,F,I,L). At higher magnifications the typically contoured surface and pores of the vitelline membrane of untreated X. laevis (D) and X. borealis (J) become apparent, scale bar = 10 m. One minute protease treatment of oocytes affects the vitelline membrane of X. laevis (E) and X. borealis (K) differently, scale bar = 10 m. Microvilli projections on the plasma membranes of X. laevis (F) and X. borealis (L) oocytes show different morphology and density, scale bar = 2 m. Each panel represents a separate oocyte.

40 The higher magnification images more clearly revealed differences in the membranes of X. laevis and X. borealis oocytes. The vitelline membrane of X. laevis (Figure 3.5D) appears to be smoother yet more porous when compared to that of X. borealis (Figure 3.5J), as evidenced by the uniform distribution of pores. There was also a notable difference between species in the effect of protease on the vitelline membrane. The membrane of X. laevis appears largely unaffected after 1 min exposure to protease (Figure 3.5E), whereas the integrity of the X. borealis membrane is obviously compromised and showed substantial wrinkling (Figure 3.5K). The structure and distribution of microvilli projections on X. laevis (Figure 3.5F) and X. borealis (Figure 3.5L) plasma membranes are also quite distinct. As previously imaged, the X. laevis microvilli show a flattened surface and appear tightly compressed (125,126). In contrast, the X. borealis microvilli are more filamentous and sparsely packed, resulting in what appears to be a more porous plasma membrane structure.

3.4 DISCUSSION

We report here the first use of X. borealis oocytes for the functional characterisation of neuronal ion channels, and identify some promising advantages over the classic X. laevis oocyte expression system. Voltage- and ligand-sensitivity, as well as exogenous channel expression levels, were very similar between the two species and comparable to literature values for all ion channels tested (summarised in Table S3.1). The pharmacological profiles of channels expressed in oocytes from each species were essentially identical in terms of efficacy and potency of small molecules agonists to larger peptide antagonists. Our results reveal that oocytes from X. borealis are a reliable substitute for X. laevis oocytes for functional and pharmacological characterisation of ion channels.

Interestingly, we observed reduced maximal current levels of the endogenous Ca2+- activated chloride channel in X. borealis oocytes. This endogenous current is one of the biggest drawbacks of studying voltage-gated ion channels in X. laevis oocytes (127). The current is observed when using high-voltage pulses and can interfere with interpretation of electrophysiological recordings particularly when studying current-voltage relationships, or channels that are substantially permeable to Ca2+. Considering an equivalent level of endogenous channel density across the oocyte membrane of both species, it is expected that the smaller X. borealis oocytes would have a smaller total Ca2+-activated chloride channel current. However we show that overexpression of exogenous channels is similar between species (Table 3.1), and therefore the ratio of exogenous to endogenous current 41 levels would nevertheless be greater for X. borealis oocytes. Thus, the significantly smaller endogenous current size in X. borealis (Figure 3.2) provides an unexpected advantage over the commonly used X. laevis oocytes.

Imaging of the surface morphology of oocytes from both species highlighted a substantial difference in the structure of the plasma membrane. We noted a dramatic increase in the fragility of the X. laevis oocytes after vitelline membrane removal as compared to the oocytes from X. borealis and suggest that the difference in plasma membrane structure could account for this observation. This difference in the membrane may impede the permeability of some molecules and indeed reveal a difference between each species oocytes’ that could be either beneficial or limiting to some studies.

Although X. laevis is the most widely used source of oocytes for ion channel studies, the oocytes of several other species have also been assessed for their suitability as a heterologous expression system for membrane proteins. Functional expression of the eel electroplax acetylcholine receptor was demonstrated in oocytes from the Japanese fire belly newt, Cynops pyrrhogaster (128). Despite the larger size (1.6–1.9 mm in diameter) and apparently more robust nature of Cynops oocytes as compared to X. laevis oocytes (129), this system does not appear to have been widely adopted for ion channel studies, likely due to limited availability. The cane toad, Bufo marinus, has also been assessed as a possible substitute source of oocytes, and was shown to robustly express KV1.1 channels (130). However, the levels of endogenous background current were slightly higher in B. marinus oocytes than those from X. laevis. This finding may explain the observation that oocytes from B. marinus have quite leaky membranes (131). Furthermore, B. marinus oocytes were less effective than those from X. laevis at expressing human amino acid transporters (131). In contrast, oocytes from the axolotl (or Mexican salamander, Ambystoma mexicanum) appear to have lower levels of endogenous ion channels. In particular they lack the Ca2+- activated chloride channel, a property that was exploited to aid expression cloning of this channel from Xenopus oocytes and characterisation of mouse orthologues (132). Despite the promising discovery of an oocyte with an electrically cleaner background, Ambystoma oocytes have not become a commonly used model system.

In conclusion, we demonstrated that oocytes from X. borealis are an excellent model system for expressing exogenous ion channels for electrophysiological studies. We also show that X. borealis oocytes have a reduced background chloride current. This is of increasing

42 importance due to the growing number of mutant channels being identified as influential in several human diseases that need be better characterised.

2.5 SUPPLEMENTARY INFORMATION

Table S3.1: Comparison of half-maximal response and slope values for ion channels tested in X. laevis and X. borealis oocytes and references to studies reporting similar values for these parameters. Steady-state desensitisation: SSD; Activation: act. There is no evidence of a difference between the two species for any parameter measured (unpaired t-test, P > 0.05).

Ion Channel Half-maximal response P value Slope P value Ref. X. laevis X. borealis X. laevis X. borealis Voltage-gated potassium channels

KV10.1 19.7 mV 20.2 mV 0.86 20.8 20.6 0.90 (113)

KV11.1 –24.3 mV –23.7 mV 0.18 6.98 7.53 0.16 (133) Voltage-gated sodium channels

NaV1.2 –22.5 mV –22.3 mV 0.61 4.74 4.60 0.84 (134)

NaV1.5 –33.5 mV –33.8 mV 0.49 4.81 4.22 0.08 (135)

NaV1.7 –14.2 mV –14.8 mV 0.09 5.24 4.88 0.27 (136) Acid-sensing ion channels

ASIC1a pHact pH 6.11 pH 6.06 0.27 1.95 1.86 0.83 (45)

ASIC1a pHSSD pH 7.19 pH 7.19 0.66 10.39 9.97 0.69 (45)

ASIC1b pHact pH 5.95 pH 5.97 0.11 3.98 3.36 0.12 (22)

ASIC1b pHSSD pH 6.78 pH 6.78 0.72 4.17 4.09 0.91 (37)

ASIC2a pHact Ambiguous fits with sigmoidal non-linear regression model

ASIC2a pHSSD pH 5.99 pH 5.98 0.66 2.27 2.37 0.78 (50)

ASIC3 pHact pH 6.16 pH 6.12 0.13 1.46 1.60 0.40 (22)

ASIC3 pHSSD pH 7.04 pH 7.02 0.19 8.25 9.04 0.70 (74) Pc1a:ASIC1a 1.00 nM 1.03 nM 0.63 1.44 1.60 0.34 (34) APETx2:ASIC3 64.7 nM 51.1 nM 0.05 0.84 0.93 0.37 (62)

43

CHAPTER 4

Defining the molecular interactions and mechanism of PcTx1 activity at rat and human ASIC1a and rat ASIC1b

44 Preface to Chapter 4. This chapter is derived from a manuscript currently under review. Conception of the work in this chapter was by Lachlan D Rash, Glenn F King, and me. Lachlan D Rash and I performed the comparison of rat and human ASIC1a (mechanism and peptide pharmacophore data). I performed the rASIC1b data. I produced peptides with Natalie J Saez and Irène R Chassagnon. Lachlan D Rash and I wrote, edited, and prepared the manuscript.

4.1 INTRODUCTION

Acid-sensing ion channels (ASICs) belong to the wider superfamily of amiloride sensitive degenerin/epithelial sodium channels (4), and open in response to low extracellular pH such as during inflammation and ischemic stroke. There are six main ASIC subtypes (1a, 1b, 2a, 2b, 3, and 4) with varied expression throughout vertebrate tissue. They assemble as functional homo- or heterotrimeric complexes presenting a broad spectrum of pharmacological, ion selectivity, and pH-dependent kinetic and steady-state properties (22,33). The ASIC1 splice isoforms ASIC1a and ASIC1b, are of interest from a therapeutic perspective in conditions associated with a sustained local extracellular acidosis. ASIC1a is the predominant isoform expressed in the central nervous system and its inhibition is potentially beneficially in several neurological disorders (e.g. ischemic stroke, spinal cord injury and multiple sclerosis) (56,137,138). In contrast, ASIC1b is expressed in peripheral neurons and its activity was recently shown to play a role in nociception in rodents (18).

Several ASIC1 modulators have been discovered ranging from small molecules to larger polypeptides, with venom-derived peptides proving the most potent and selective compounds to date (139). The best studied is the 40-residue peptide PcTx1 (also -TRTX- Pc1a), isolated from the venom of the spider Psalmopoeus cambridgei (34). PcTx1 is an inhibitor of rodent ASIC1a homomers (IC50 1 nM) and ASIC1a containing heteromers

(ASIC1a/2b and ASIC1a/2a, IC50 ~3 nM) (35,36,140), whilst potentiating ASIC1b homomers

(EC50 ~100 nM) (37). PcTx1 acts as a gating modifier by binding to the functionally important acidic pocket of ASIC1 (48). The PcTx1:ASIC1 complex structure revealed three peptides bound per channel at the subunit interface, with an extensive network of intermolecular contacts with both the primary (thumb domain) and complementary (palm and beta ball) face of the acidic pocket (27). The functional pharmacophore of PcTx1 at rat ASIC1a (rASIC1a) was demonstrated by alanine-scanning mutagenesis of PcTx1 to comprise Trp7, Trp24, Arg26, Arg28, and Phe30 (41,51).

45 Binding of PcTx1 to rASIC1a leads to an increase in the apparent proton affinity of channel gating, shifting the steady-state desensitization (SSD) and activation curves to more alkaline values (48). Under most circumstances this renders channels desensitised and unable to open in response to protons, however if desensitisation is not achieved, pH-dependent activation is enhanced. This mechanism of action was recently shown to be shared for heteromeric rASIC1a/2a channels, and proposed to be a more general mechanism by which PcTx1 modulates ASICs (140). Evidence for this same mechanism at human ASIC1a (hASIC1a) is supported by the finding that PcTx1 inhibits hASIC1a when applied at pH 7.2 but not at 7.4 (49). }. Surprisingly, this species-dependent PcTx1 activity has only been studied with whole P. cambridgei venom (at 1:10,000 dilution), which has previously been shown to be highly complex cocktail of molecules, with PcTx1 contributing <0.5% of the total venom (57). This confounds whether other molecules within the venom have activity at either rat and/or human ASIC1a. In addition, neither a full steady-state desensitisation curve or concentration-response curve under physiological conditions has ever been reported with PcTx1 at hASIC1a. Previous studies have only tested ASIC1a species differences at a single concentration, and found contrasting results at high application pH.

The PcTx1 mechanism of action for potentiation of rASIC1b is due to a shift in the activation curve to more alkaline values (with little change to the pH-dependence of steady-state desensitisation), and a slowing of the desensitisation kinetics of proton-induced currents (37). The different functional outcomes of ASIC1 species and subtype variants upon PcTx1 exposure is somewhat surprising given the high level of sequence identity between these channels. Rat and human ASIC1a differ by only eight point mutations, and a two-residue insert in the human channel. Importantly, the only surface-exposed residue in the PcTx1 binding region that differs between these channels is Ala178 in rASIC1a, which is a valine in hASIC1a. Rat ASIC1a and 1b are splice isoforms that are identical in the second third of the protein (from residue 186 of rASIC1a onwards). Only a small portion of the overall PcTx1 binding region (<30% of crystal contacts) consists of residues before the ASIC1a/1b splice site, with crystal contacts observed for residues between 174 and 179 of rASIC1a (27). In both cases, the differences between channels are present on the complementary face of the PcTx1 binding region.

Here we present an in-depth analysis of the mechanistic basis for the difference in PcTx1 potency at human and rat ASIC1a species variants. In addition to the inherent steady-state desensitisation characteristics of a channel that give rise to this potency difference, we also

46 probe the molecular interactions of PcTx1 with hASIC1a via alanine scanning mutagenesis of the peptide. This scan was also performed at rASIC1b and identified potential residues to target to in order to manipulate PcTx1 subtype selectivity. These insights contribute significantly to our understanding of the molecular details of the interactions between PcTx1 and ASIC1, and could assist in developing PcTx1 or similar peptides as molecular probes and therapeutic lead molecules.

4.2 METHODS

4.2.1 Data analyses

Concentration-response curves were compared as logIC50 values with an extra sum-of- squares F test (two-groups) or one-way ANOVA with Bonferroni correction for multiple comparisons (> two groups); in each case assuming a Gaussian distribution and equal standard deviations. Data with a single concentration were compared with a one-way ANOVA with Dunnett’s test for comparison to a particular mean (wild-type value). Intercepts of linear regressions (Schild plot) were compared with an Analysis of Covariance (ANCOVA). All data points are shown as mean ± SEM (standard error of the mean), and P < 0.01 was considered to be significant for all data.

4.3 RESULTS

4.3.1 PcTx1 is less potent against human ASIC1a than the rat isoform

Using pure recombinant peptide, we found that PcTx1 applied at pH 7.45 inhibits hASIC1a expressed in Xenopus oocytes with an IC50 of 3.2 nM, or 10-fold higher compared to its inhibition of rASIC1a under the same conditions (Figure 4.1B and C). This is consistent with a previous report that the same dilution of PcTx1 venom is substantially less effective at hASIC1a than the mouse isoform (49). In order to assess if this was due to species differences in the binding site, we tested PcTx1 activity at the “humanised” rASIC1a A178V mutant. This substitution is the only surface-exposed residue in the PcTx1 binding region of the channel that differs between rat and human ASIC1a (Figure 4.1A). PcTx1 was equipotent against wild-type rASIC1a and the A178V mutant (Figure 4.1B). Having discounted a direct interaction as the basis for the 10-fold difference in potency of PcTx1 against rat and human ASIC1a, we analysed the dependence of conditioning pH (via steady- state desensitisation curves) on PcTx1-mediated inhibition of ASIC1a. Desensitisation of the channel is a process linked closely with inhibition because if all channels become

47 desensitised, they are no longer available to open and conduct current. Figure 4.1D shows the steady-state desensitisation curves for rASIC1a, the humanised rASIC1a A178V, and hASIC1a. Desensitisation of hASIC1a is significantly less sensitive to protons than rASIC1a (P < 0.0001) and that introducing an A178V mutation in rASIC1a has no effect on the pH- dependence of SSD (P = 0.7098).

Figure 4.1: Comparison of PcTx1 activity at rat and human ASIC1a. (A) Structure of cASIC1 with PcTx1 bound at the acidic pocket and the position of residue 178 highlighted in red (PDB code 4FZ0). (B) Example traces for PcTx1 concentration-response (left) and steady-state desensitisation (right) curves for rat (black) and human (blue) ASIC1a. (C) Concentration-response and (D) Steady-state desensitisation curves for rASIC1a, hASIC1a, and the humanised A178V mutant of rASIC1a. Data points represent mean ± SEM (n = 4–7).

We then assessed the effect of four different concentrations of PcTx1 on the pH- dependence of steady-state desensitisation of rat and human ASIC1a, as well the rASIC1a A178V mutant (Figure 4.2A–C and Table 4.1). Increasing the PcTx1 concentration increased the alkaline shift in pH50, with the degree of shift being relatively similar for all three channels at a given PcTx1 concentration. To better assess the relative affinity of PcTx1 for each channel this data was used to construct a Schild Plot. Specifically, the pH50 ratios (in the absence and presence of peptide) for each concentration of PcTx1 were used and the pA2 interpolated from the abscissa intercept of the graph (Figure 4.4D). This analysis revealed pA2 values of 8.43, 8.49 and 8.60 for rASIC1a, rASIC1a A178V, and hASIC1a,

48 respectively. A pA2 value represents the negative log of the concentration of antagonist that results in a two-fold shift in the agonist response curve and represents the affinity of the antagonist for the receptor. The values obtained here for rASIC1a, rASIC1a A178V, and hASIC1a correspond to PcTx1 concentrations of 3.72, 3.21 and 2.54 nM, respectively. The similarity of the pA2 values between the channels suggests that there is essentially no difference in the affinity of PcTx1 for the channels (P = 0.8474). This further shows that the difference in the proton sensitivity of steady-state desensitisation between the rat and human channel explains the apparent species-dependent difference in potency of PcTx1.

Table 4.1: Effect of increasing concentrations of PcTx1 on steady-state desensitisation1.

Condition Rat WT Human WT Rat A178V Control 7.329–7.337 7.038–7.046 7.322–7.349 0.35 nM PcTx1 7.453–7.464 7.119–7.143 7.441–7.468 1 nM PcTx1 7.508–7.533 7.212–7.244 7.536–7.565 3.5 nM PcTx1 7.632–7.656 7.411–7.453 7.649–7.678 10 nM PcTx1 7.686–7.759 7.529–7.554 7.705–7.752

1 Shown are 95% Confidence Intervals of pH50 values.

The PcTx1-mediated functional effect and on rate at rASIC1a are pH dependent (Figure 4.2E). PcTx1 application for 5 s at pH 7.45 and above resulted in potentiation of currents, while at pH 7.3 inhibition was observed. This demonstrates that under all conditions, the interaction and functional effect of PcTx1 is extremely rapid. For conditioning pH 7.75 (an alkaline shift of 0.42 units from pH50 of steady-state desensitisation), no inhibition was observed even with constant perfusion of 3 nM PcTx1 for 10 minutes. The recovery from PcTx1 activity is also pH dependent, with a faster off-rate when conditioning at higher pH (Figure 4.2F).

49

Figure 4.2: pH-dependent effects of PcTx1 for ASIC1a. Effect of increasing concentrations of PcTx1 on the steady-state desensitisation of (A) rASIC1a, (B) hASIC1a, and (C) humanized A178V mutant of rASIC1a. Control currents were taken from pH 7.45 to pH 6, with all data normalized to this peak. (D) Schild plot for inhibition of ASIC1a showing no significant difference in y-intercepts for each channel (P = 0.8474, F = 0.169, DFn = 2, DFd = 8). (E) Time course of rASIC1a inhibition by 3 nM PcTx1 with variable conditioning pH. Right panel is zoomed into the first 30s of the on-rate graph. (F) Recovery of rASIC1a from inhibition by 3 nM PcTx1 (110 s application) at different conditioning pH. Data points represent mean ± SEM (n = 4–7).

Given the pH dependence of on- and off-rate, we next assessed the potency of PcTx1 when applied at different conditioning pHs (Figure 4.3). As expected from the SSD data in Figure 4.2, application of peptide at more acidic conditioning pH results in more potent activity for all channels. Comparison of rat and human IC50 values at a given shift in pH from the pH50

SSD for each channel reveals equipotent activity in agreement with the Schild plot. As the conditioning pH becomes more alkaline, there is a point at which PcTx1 is no longer an inhibitor of channels. With a further increase in the conditioning pH, PcTx1 indeed potentiates ASIC1a currents in a concentration-dependent manner in agreement with the on-rate studies at pH 7.75.

50

Figure 4.3: Activity of PcTx1 at different conditioning pHs. Example traces of PcTx1 concentration- response data for rat (top) and human (bottom) ASIC1a conditioned at pH50 SSD + 0.26–0.27 pH units. Data for (A) rASIC1a, (B) hASIC1a, and (C) humanized A178V mutant of rASIC1a. Data points represent mean ± SEM (n = 4–7).

4.3.2 Pharmacophore of PcTx1 against human ASIC1a

To further study potential differences between rat and human ASIC1a modulation by PcTx1, we performed an alanine scan at hASIC1a to determine the PcTx1 functional pharmacophore (see Table 4.2 for species comparison). The PcTx1 binding surface for activity at both channels is very similar, with mutation of the key pharmacophore residues as determined on rASIC1a (41,51) also resulting in a loss of inhibition at hASIC1a (Figure 4.4). However, despite the near identity of the binding site between the species isoforms, some differences were observed (Figure 4.4F). Deletion of two N-terminal residues of PcTx1 (N2) resulted in a mild loss of activity of at the human channel but had no effect on activity at the rat channel, suggesting the N-terminus is more important to hASIC1a activity (Figure

4.4F). Mutation of K25 to an alanine had little effect on the IC50 for human and rat ASIC1a, however the inhibition was never complete at hASIC1a with <75% inhibition even at saturating concentrations, whereas it retained full efficacy at the rat channel (Figure 4.4G). The R27A mutant lost a similar degree of potency at both rat and human ASIC1a. However,

51 similar to the K25A mutant, the efficacy of R27A at hASIC1a was incomplete and saturated at ~80% inhibition, whilst it retained full efficacy on the rat channel (Figure 4.4H).

Table 4.2: Summary of wild-type and mutant PcTx1 IC50 values on rat and human ASIC1a

Human Rat1

2 2 IC50 (nM) Fold change IC50 (nM) Fold change WT 3.14 – 0.35 –

N2 22.4 7.1 0.47 1.3 K6A ~2790 888 34.2 97 W7A >1000 >318 >1000 >2857 K8A 5.23 1.7 0.59 1.7 H14A 4.31 1.4 0.26 0.7 W24A >1000 >318 72.0 205 K25A 4.98 1.6 0.64 1.8 R26A 164 N/A 1880 N/A R27A 33.1 10.5 3.01 8.6 R28A 106 33.8 5.06 14.5 S29A 3.30 1.1 0.32 1.1 F30A 25.6 N/A 6.79 N/A E31A 3.00 1.0 0.76 2.2 V32A 63.3 20.2 6.27 18 V34A 2.09 0.7 0.49 1.4 P35A 4.24 1.4 0.13 2.7 K36A 4.99 1.6 0.19 1.8 T37A 7.69 2.5 0.25 1.4 P38A 10.8 3.4 1.36 3.9

C3 8.77 2.8 1.44 4.1

1Values for rat ASIC1a not from this study are from (41,51). 2Fold change in activity is relative to WT PcTx1 for each species variant. Italicized numbers indicate an increase in potency, and bold a difference between rat and human activity.

52

Figure 4.4: Structure-activity relationship of PcTx1 at hASIC1a. (A-D) Concentration-response curves for modulation of human ASIC1a by WT and mutant PcTx1. (E) Mean ± SEM pIC50 values.

Dashed line represents WT pIC50 value and the grey background is a half-log unit either side of the pIC50. Bold and italicised mutants represent a loss of activity compared to WT (one-way ANOVA with Dunnett’s test: F = 152.4, p < 0.001). (F-H) Comparison of rat and human concentration-response curves for PcTx1 WT, N2, and K25A. Data points represent mean ± SEM (n = 3–6).

4.3.3 Pharmacophore of PcTx1 at rat ASIC1b

We also assessed the activity of our series of PcTx1 alanine mutants at rASIC1b in order to determine if the pharmacophore residues were the same as at rASIC1a. PcTx1 is a potent potentiator of rASIC1b with an EC50 of 24.6 nM, ~30-fold higher than the rASIC1a IC50 (Figure 4.5A). The ability of each PcTx1 mutant to potentiate rASIC1b was assessed using two concentrations selected from the slope of the wild-type peptide concentration-response

53 curve and compared to the relative change in activity of these mutants at rASIC1a (Figure 4.5B–D). Mutations of all the primary rASIC1a pharmacophore residues had an equally deleterious effect on PcTx1’s activity at rASIC1b. In contrast, alanine mutations to Lys25 and Thr37, which had minimal effect on rASIC1a activity, substantially decreased activity at rASIC1b. Conversely, the C3 truncation showed a loss of rASIC1a activity with no change to the potentiation of rASIC1b. The potentially improved selectivity of PcTx1 K25A and T37A mutants towards rASIC1a was confirmed upon obtaining full concentration-response curves which revealed apparent effects on both the efficacy and potency at rASIC1b (Figure 4.6B and C). In addition to the individual mutations, a double mutant (K25A/T37A) showed a further decrease in rASIC1b activity with little effect on rASIC1a potency, thus has ~3 to 4- fold improved rASIC1a selectivity compared to wild-type peptide. Finally, we also assessed the effect of mutating residues Trp24 and Arg28 as they are key pharmacophore residues at rASIC1a for PcTx1, but substituted in the homologous peptide Hi1a (Figure 4.6A), a subnanomolar inhibitor of rASIC1a with substantial similarity to PcTx1 yet causes only minimal potentiation of rASIC1b (~130% of control at 1 μM) (43). Unfortunately none of the tested mutations to residues 24 and 28 of PcTx1 improved ASIC1a selectivity, while also retaining suitable potency (Figure 4.6B and C). Intriguingly, the W24F and W24Y mutants showed an incomplete block at rASIC1a at saturating concentrations, similar to what is observed for Hi1a (contains a Tyr at position 24).

54

Figure 4.5: Comparative activity of PcTx1 mutants at rASIC1a and rASIC1b. (A) Concentration- response curves for wild-type PcTx1 at rASIC1a and rASIC1b (B) Example traces for PcTx1 WT, K25A, and T37A peptides at rat ASIC1a and 1b. (C and D) Relative activity of PcTx1 analogues compared to wild-type PcTx1 for inhibition of rASIC1a (red circles) and potentiation of rASIC1b (black squares) at pH 7.45 conditioning. Grey shading indicates mutants where the change in activity was different between ASIC1a and ASIC1b for that variant. rASIC1a activity for the R26A and F30A mutants are not shown as they potentiate, and not inhibit, this channel. Data points represent mean ± SEM (n = 3–5).

55

Figure 4.6: Subtype specificity of PcTx1 analogues. (A) Amino acid sequence alignment of the ASIC1 modulating peptides PcTx1, Hm3a, and Hi1a. Differences in sequences are indicated by grey boxes, and pharmacophore residues for PcTx1 activity at rASIC1a by red dots above the residue. Concentration-response curves for PcTx1 mutants at (B) rASIC1a and (C) rASIC1b. Data points represent mean ± SEM (n = 3–5).

4.4 DISCUSSION

The tarantula peptide PcTx1 was the first potent and selective ASIC1 modulator discovered (34), making it one of the most widely used tools to study ASIC function in vitro and in vivo. Nevertheless, the details of how it interacts with several ASIC1 variants to produce different species and subtype-dependent activities is far from clear. Here we show that despite highly similar binding sites on rASIC1a, 1b and hASIC1a, PcTx1 can have subtly (rat and human ASIC1a) or substantially (rat ASIC1a and 1b) different functional consequences on these channels, and that mutations of PcTx1 that face the complementary binding surface (i.e. position Lys25) may be able to be exploited to modify the species and subtype selectivity of the peptide.

56 A recent proposal for the mechanism of ASIC gating involves a linear activation model with channel opening requiring multiple protons to bind to the resting state, and desensitisation arising exclusively from the open state (21). During proton-dependent activation the sequential protonation events brings carboxylate pairs between the thumb and finger/palm domains closer together. With each protonation, this pinching motion is made more energetically favourable (high cooperativity of gating) as it draws unprotonated neighbouring residues closer together (29,30,141). This gating rearrangement is consistent with structural studies that revealed an expanded acidic pocket in the resting state compared to that of the open and desensitised states (28). In the continued presence of protons, ASICs undergo desensitisation. This provides a technical limitation, as if the kinetics of desensitisation are faster than the application of protons to all channels being studied, the overlap of ASIC desensitisation will likely mask the true extent of proton-induced activation. These technical limitations can be somewhat overcome by the use of macropatch recordings and fast perfusion systems which can measure ASIC deactivation has been observed opposed to desensitisation. This has been shown in two studies with the findings agreeing with the linear activation model described above (43,141). ASICs also undergo steady-state desensitisation when channels are subjected to mildly acidic conditions (between pH 7.4– 6.9) that do not induce an observable current, however transition the channel to a desensitised state that can no longer activate in response to low pH-stimulus.

The majority of experimental data published on ASIC modulators has involved the use of rodent channels. To our knowledge this is the first report of a concentration-response curve for PcTx1 activity at hASIC1a under physiological conditions (pH ~7.45), where we show a 10-fold decrease in activity at the human over rat channel. Species dependent potency of PcTx1 and other ASIC modulators has implications for their use and development as pharmacological tools and therapeutic candidates. In the case of PcTx1 the species- dependent potency has little to do with the point mutation at residue 178, despite this being the only change to the PcTx1 binding region between rat and human ASIC1a. Therefore this mutation is insufficient to cause a change in the interaction and functional consequence of PcTx1 binding and suggests that the observed species difference in potency is due to the mechanism of action of the peptide as opposed to binding interaction.

PcTx1 inhibits rat ASIC1a via shifting the steady-state desensitisation curve to more alkaline values rendering channels desensitised at physiological pH (48), however some aspects of its mechanism of action and binding mode are unclear. This mechanism of action is the

57 reason for the 10-fold difference in potency we observed between human and rat ASIC1a.

Specifically, when applied at pH 7.45 (or the pH50 SSD + 0.12) in order to produce 50% desensitisation of rASIC1a, PcTx1 only needs to cause a shift of 0.1 pH units in the alkaline direction, whereas under the same conditions, for hASIC1a (pH50 SSD 7.04) this would require a shift of 0.41 pH units, and consequently a higher concentration of PcTx1. Our data extends upon the preliminary work by Sherwood and Askwith who tested the activity of P. cambridgei venom (at 1:10,000 dilution) at pH 7.2 and 7.4 at both human and mouse ASIC1a, and only observed inhibition of hASIC1a at pH 7.2 (49). The use of whole venom in this study confounds whether other molecules within the venom have activity at either rat and/or human ASIC1a, as this venom has been shown to be a mixture of molecules, with PcTx1 contributing <0.5% of the total venom (57). Sherwood and Askwith also showed the hASIC1a AKPDL mutant (A274S/P285S/K291N/D298/L299 to mimic rASIC1a) recapitulated both the PcTx1-induced (using P. cambridgei venom) inhibition at pH 7.4 and pH-dependent steady-state gating properties of wild type mASIC1a (49). Together this shows that the apparent potency of PcTx1 is dependent on the pH at which it is applied — the closer to the pH50 of steady-state desensitisation, the higher the apparent potency for inhibition.

The reported effect of PcTx1 at rat and mouse ASIC1a when preconditioned at high pH (7.9 or 8.0) is contradictory, with some studies finding concentration-dependent inhibition (36,53) and others showing either no observable activity or potentiation of proton-gated currents (48). By performing full SSD curves for both rat and human ASIC1a in the presence of varying concentrations of PcTx1, we have demonstrated that PcTx1 is a concentration- dependent potentiator of pH 6-evoked currents when applied at high pH (>7.75). Consistent with this we see no inhibition at pH 7.75 even with 10 minutes of application, which indicates that PcTx1 is unable to ‘induce’ the desensitised state under these conditions. Specifically, the peptide alone cannot force the channel to undergo conformational rearrangement from the high pH resting state (without protons bound) to the non-conducting desensitised state. When PcTx1 is applied at high pH the positively charged side chains of the peptide may act in a similar manner to protons and interact with ASICs to assist in the cooperativity of pH- dependent gating, but not enough that channels enter a desensitised state, ultimately resulting in enhanced low pH evoked currents. It is also important to consider the occupancy of the three PcTx1 binding sites on the channel. For example it may be the case that at high pH where the binding affinity is low (53,) at high enough concentrations PcTx1 binds the majority of channels but with only partial occupancy which gives rise to potentiation. This is 58 in agreement with the potentiation effect at rASIC1a being observed at PcTx1 concentrations of >3 nM, whereas inhibition is achieved at ~30-fold lower concentrations when in the presence of co-operating protons. This is further supported by our results showing a slower on- and off- rates at high pH, consistent with fewer peptide channel interactions taking place.

Our data suggests that at pH 7.6 and below, enough protonation events have occurred that rASIC1a is in a conformation/state of surface charge where the PcTx1 off rate is sufficiently slowed that more binding sites per channel are occupied at a given time to stabilise the desensitised state giving rise to inhibition of current. The more protons bound, the more favourable the channel conformation for PcTx1 binding as it is structurally closer to the desensitised state, thus the higher inhibitory potency (see Figure 4.3). This is in agreement with early I125-PcTx1 binding studies to rat brain and rASIC1a overexpressed in CHO cells (53). This binding data showed a bell curve of binding that was maximal at ~pH 7, with >10% of maximal binding occurring between pH 5.75 and 7.75. The observations of initial potentiation of currents with 5 second PcTx1 application at pH 7.45 and 7.6 as seen in Figure 3E supports our suggestion that partial occupancy of the three PcTx1 binding sites results in potentiation, while higher binding site occupancy results in channel desensitisation and functional inhibition.

PcTx1 has also been shown to stabilise the open state of ASIC1a, however with lower affinity than for the desensitised state. Binding to the open state was initially shown by application of PcTx1 at pH 7.1 inducing rASIC1a currents, whereas pH 7.1 alone does not open channels (37). A concentration-response curve using this application protocol revealed a lower limit for the EC50 of 156 nM. The true EC50 for the open state is hard to determine, as a plateau in maximal response was not reached, as well as current desensitisation likely masking complete activation. Nevertheless, this is in agreement with the notion that PcTx1 can function in a similar manner to protons to modulate ASICs, and that the binding event can be rapid. Although studying different steps in the linear kinetic model, together these studies suggest PcTx1 can bind multiple ASIC1a states, however binding to the desensitised state is preferred.

Although the ability of PcTx1 to act on ASIC1a in a similar manner to protons can explain the difference in observed potency of PcTx1 for human and rat ASIC1a, our PcTx1 alanine scan showed that there are subtle difference in the pharmacophore residues for each species variant. Given the significant state-dependent rearrangements of the acidic pocket 59 where PcTx1 binds (25,26,28), it is conceivable that PcTx1 makes subtly different peptide:channel interactions to what has been shown with electrophysiology at pH 7.45 (41,51) as well as in the co-crystal structures where the extracellular domain is most similar to the desensitised state (27,52). Species-dependent differences in activity of PcTx1 N2, K25A, and R27A variants could be an indication that these regions play different roles in the overall binding kinetics/mechanism of PcTx1 to these isoforms of ASIC1a. It is intriguing to note that in the PcTx1:ASIC1 structure, Lys25 and Arg27 of PcTx1 made crystal contacts almost exclusively with the complementary face (13 of 14 contacts combined) of the acidic pocket where the single A178V mutation lies. This highlights the potential to perform further mutagenesis at these positions in PcTx1 in an attempt to further alter the species dependent ASIC1a activity. Given the conditioning pH was kept constant at 7.45, both species variants would present in the resting state, however at a different stage on the linear model with a different amount of protonation sites occupied.

PcTx1 is widely used as a selective ASIC1a inhibitor, however the significant rASIC1b potentiating activity (Figure 4.5A) may confound any observed findings with PcTx1. Thus having more subtype selective tools would be of great value. The extensive overlap in pharmacophore across rat ASIC1a and 1b was of some surprise given the different functional consequence of PcTx1 application at these subtype variants (37,48). However, the overall orientation of the PcTx1 binding region of ASIC1 is extremely similar in the desensitised and open state structures giving credence to our results (25,26). The functional importance of channel residues that differ between rASIC1a and 1b that are both pre-splice site and crystal contacts have been studied for Hm3a (82% identity with PcTx1) (60). Within this region, the rASIC1a mutants R175C and E177G (rASIC1a to 1b substitutions) caused a 10–20-fold loss in inhibitory activity by Hm3a. These residues on the complementary surface are indeed in close proximity with PcTx1 residues 24–28. Here we produced a double mutant analogue of PcTx1 (K25A/T37A) with slightly improved selectivity for rASIC1a inhibition over rASIC1b potentiation compared to the wild-type peptide. The W24Y and R28H mutants (PcTx1 to Hi1a mutations) did not improve selectivity, suggesting either the second half of Hi1a, or the different mechanism of action and binding orientation of the double knot peptide compared to PcTx1 is responsible for its more pronounced ASIC1a selectivity. The incomplete inhibition of PcTx1 W24F and W24Y at rASIC1a (similar to Hi1a) could be explained by interactions from this residue destabilising the open-to-desensitised gating step. Although we were unable to design any truly species or subtype selective PcTx1 analogues, this work highlighted that residues interacting with the complementary face of 60 ASIC1 can likely be targeted for further mutagenesis to develop more selective PcTx1 analogues.

In summary, we show that the different apparent potency of PcTx1 on rat and human ASIC1a is due to variations in their steady-state desensitisation profiles, rather than differences in the PcTx1 binding affinity or mutation in the ASIC1a acidic pocket binding region. The functional outcome observed depending on PcTx1 application pH is an important point for consideration when using the peptide as a pharmacological tool to study ASICs particularly in vivo where the exact pH under which the peptide is acting is not known. Furthermore, alanine-scanning mutagenesis revealed subtle differences in the PcTx1 pharmacophore at rASIC1a, hASIC1a, and rASIC1b. It will be interesting in future studies to further develop PcTx1 analogues to be more selective for a specific ASIC variant and activity, which will add to the toolkit of pharmacological modulators available to understand ASIC function. Together, our findings have furthered our understanding of the molecular details governing the interaction between PcTx1 and ASIC1 variants.

61

CHAPTER 5

APETx2 mechanism of action, binding site, and rat ASIC1b pharmacophore

62 5.1 INTRODUCTION

Pain is a complex disease with many current therapies lacking sufficient efficacy or patient tolerability. This has led to a sustained research effort focussed on understanding nociceptive signalling pathways, and identifying ion channels that have emerged as targets for developing future analgesics (142). Pain is often associated with tissue acidosis, where a drop in local extracellular pH can initiate firing of sensory neurons (143,144). Acid-sensing ion channels (ASICs) are activated by this drop in extracellular pH and can function as the primary proton sensors in neurons (6). Among the several ASIC subtypes, ASIC1a, 1b, and 3 have been shown to be expressed in peripheral nociceptors of rodents and humans, making these channels prime therapeutic candidates for treatment of peripheral pain (18,39,70).

Pharmacological modulators have proven to be the most effective research tools in revealing the involvement of ASICs in pain (13,33). The snake toxin MitTx (ASIC1a, 1b, and 3 agonist) and synthetic molecule GMQ (ASIC3 agonist) elicit robust pain-related behaviours upon intraplantar injection in mice, effects that are ablated in ASIC1-/- and ASIC3-/- mice, respectively (76,145). Similarly the conopeptide RPRFa is a potentiator of low pH-induced ASIC3 currents in vitro, and enhanced acid-induced muscle pain in wild type but not ASIC3- /- mice (95). In agreement with this work, the ASIC3 inhibitor APETx2 is analgesic in rodent models of chemically-induced, inflammatory, and postoperative pain (15,63,72,146). Thus, ASIC3 is the most validated of this channel family as a potential pain target.

Several ASIC3 inhibitors have recently been discovered (see Chapter 1 for overview), however APETx2 is still the most potent inhibitor and promising therapeutic candidate. APETx2 is a 42-residue peptide isolated from the sea anemone Anthopleura elegantissima (62). It was the first potent inhibitor of ASIC3-containing channels to be described, with an

IC50 of 63 nM for rat ASIC3 homomers and 0.1–2 μM for ASIC3-containing heteromers. It is also a potentiator of rat ASIC1b and 2a in the micromolar range (146). The promiscuity of APETx2 is not limited to ASICs, with inhibitory activity described at the voltage-gated channels hERG, NaV1.2 and NaV1.8 (42,65,66). The inhibition of NaV1.8 may be a beneficial off target activity as this channel has also been implicated in pain perception.

The APETx2 pharmacophore that mediates its interaction with ASIC3 was shown to comprise a largely hydrophobic patch across loops 2 and 4 of the peptide and the N-terminus (42,67). Much of this pharmacophore is shared for inhibition of hERG channels, however

63 APETx2 P18A reduced hERG activity whilst retaining rASIC3 inhibition, therefore representing a potentially improved APETx2 analogue. The peptide binding site at ASIC3 has been proposed to be either the acidic pocket (31) or a region above the transmembrane domain between the wrist and lower palm domains (68). Despite the comprehensive structure-activity studies of APETx2, its mechanism of action and ASIC binding site have remained elusive with no functional studies providing evidence for the proposed hypotheses. Here we aim to fill this gap in knowledge of an important ASIC research tool, and also reveal subtle difference in the pharmacophore of APETx2 for inhibition of ASIC3 and potentiation of ASIC1b. We propose that APETx2 binds to a novel region for peptide modulators of ASICs and acts by preventing the closed-to-open gating transition of rat ASIC3.

5.2 MATERIALS AND METHODS

5.2.1 Data analyses

Concentration-response curves were compared as logIC50 values with an extra sum-of- squares F test assuming a Gaussian distribution. Data with a single concentration were compared with a one-way ANOVA with Tukey test for multiple comparisons. On-rate kinetics were fit with a one phase exponential decay function. All data points are shown as mean ± SEM (standard error of the mean).

5.2.2 Peptides Recombinant APETx2 and PcTx1 were produced as described in Chapter 2, and MitTx was purchased from Alomone Labs (Jerusalem, Israel).

5.3 RESULTS

5.3.1 APETx2 mechanism of action

Using two-electrode voltage clamp electrophysiology the effect of APETx2 on the pH- dependence of activation and steady-state desensitisation (SSD) of rat ASIC3 (rASIC3) was examined at two concentrations. APETx2 does not shift the pH50 of activation (control = 6.16, 300 nM APETx2 = 6.10, 3 M APETx2 = 6.24; P = 0.4121; Figure 5.1A) and induces a pH-independent inhibition of the channel when applied only in the conditioning solution to the closed state (P = 0.1457; Figure 5.1B). APETx2 produced a slight concentration- dependent shift in the SSD curve to more alkaline values (control = 7.14, 100 nM APETx2 = 7.17, 1 M APETx2 = 7.22; Figure 5.1C). 100 nM APETx2 inhibition was greater with

64 decreased conditioning pH in this protocol where peptide was applied for ~2 min (Figure 5.1D). Although the number of APETx2 molecules remained consistent, an increasing population of channels had undergone steady-state desensitisation (as seen on the control curve of Figure 5.1C). This ultimately increases the relative proportion of APETx2 to closed state channels resulting in the apparent higher degree of inhibition at lower pH. This suggests APETx2 inhibits ASIC3 by binding to the closed state and prevents the closed-to- open gating transition that is normally induced by acid stimulus. The onset of APETx2 induced inhibition of rASIC3 was rapid and concentration-dependent, with a t1/2 of 13.97 s and 4.45 s for 100 nM and 1 M, respectively (Figure 5.1E). It is important to note that full activity at 100 nM is observed after ~30 s application, which is a much shorter time frame compared to the SSD data in Figure 5.1C and D, further corroborating the idea that APETx2 binds the ASIC3 closed state. Over four different holding potentials between -90 and 0 mV, the inhibition of rASIC3 by APETx2 (at both 300 nM and 3 M) was not voltage-dependent (Figure 5.1F), suggesting binding of APETx2 outside the membrane electric field. Calcium ions have been shown to stabilise multiple states of ASICs by binding at the acidic pocket, central vestibule and pore region (24,147,148). We examined the influence of calcium on the concentration-dependence of APETx2 inhibition of rASIC3, and found no difference in the IC50 when obtained in 0.1, 1.0, and 1.8 mM extracellular calcium, however a ~2 fold shift to a higher IC50 occurred in the presence of 10 mM calcium (Figure 5.1G). Although a very minor effect, this could suggest either a direct overlap in the binding site of APETx2 and calcium, or that the presence of calcium bound to the channel allosterically prevents the ability of APETx2 to exert its inhibitory effect.

65 Figure 5.1: APETx2 mechanism of action at rat ASIC3. (A) APETx2 does not shift the rASIC3 pH- dependence of activation. (B) 300 nM APETx2 is a stimulating pH-independent inhibitor. (C) Effect of APETx2 on steady-state desensitization, and (D) conditioning pH-dependent degree of 100 nM APETx2 inhibition (Comparison to pH 7.9; n.s. is not significant, and * indicates P<0.0001). APETx2 (E) On-rate, (F) voltage-dependence (300 nM: P = 0.6493, and 3 M: P = 0.0596) and (G) calcium- dependence (0.1–10 mM) of ASIC3 inhibition. Data points represent mean ± SEM (n ≥ 5).

5.3.2 ASIC3 mutagenesis for the APETx2 binding site

Docking studies of APETx2 using a homology model of rASIC3 revealed two potential binding sites, the top pocket and acidic pocket (Figure 5.2A; HADDOCK docking poses not shown) (149,150). In order to determine if either prediction was correct, a series of channel mutants to disrupt the predicted side-chain/side-chain interactions were generated. All mutant channels produced were functional, elicited currents of similar magnitude to wild- type rASIC3, and showed pH-dependence of activation properties similar to wild-type channel (data not shown); thus, any change observed in IC50 for inhibition of mutant channels by APETx2 is likely to be the direct result of an altered interaction between peptide and channel. Unfortunately, concentration-effect curves obtained for inhibition of each mutant channel by APETx2 revealed no significant difference in effect from that observed with wild-type channel (Figure 5.2B and C). Amino acids along the lower palm and thumb domain of ASIC3 were also mutated and APETx2 activity tested. These residues were selected as they are in regions involved in gating and near the fenestration that allows for entrance of ions that conduct upon proton-activation, furthermore many are crystal contacts for MitTx binding to ASIC1 (Figure 5.2A). Unfortunately, no difference in effect was observed from wild-type (Figure 5.2D).

Given the failure of the initial approach of predicting the binding site and producing mutants of surface exposed residues in these regions, a new approach was tried. In order to cover a larger region of channel in a single experiment, a series of ASIC3 chimeras were generated (Figure 5.2A and E). Substitution of the rASIC3 beta-ball, palm, and thumb regions individually with that of rASIC1a (insensitive to APETx2) did not change inhibition by APETx2, suggesting binding outside of these regions. However, substituting the entire ASIC3 extracellular domain with that of ASIC2a did completely abolish APETx2 activity. These mutagenesis studies together show that APETx2 is unlikely to bind to the acidic pocket, top pocket, or to the thumb, beta-ball, or palm domains to exert its inhibitory activity.

66

Figure 5.2: APETx2 binding site via ASIC3 mutagenesis. (A) The resting state chicken ASIC1a structure with areas mutated in this study highlighted (PDB code 5WKU). Concentration-response curve of APETx2 at ASIC3 single mutant channels in the (B) acidic pocket (C) top pocket (D) lower palm and thumb domains. (E) Activity of APETx2 at ASIC3 chimeras containing segments of ASIC1a or 2a. Data points represent mean ± SEM (n ≥ 5).

5.3.3 Functional competition assays of APETx2

To confirm the findings of the mutagenesis work, functional competition studies of APETx2 were performed with PcTx1 (at rASIC1b) and MitTx (at rASIC3). PcTx1 inhibits ASIC1a and potentiates ASIC1b, in both cases via binding into the acidic pocket (Figure 5.3A). APETx2 (10 M) and PcTx1 (100 nM) both potentiate rASIC1b to varying degrees (Figure 5.3B and C). Interestingly, co-application of these two peptides induced a sustained opening of

67 channels (Figure 5.3B and D). Application of APETx2 alone, followed by APETx2 and PcTx1 in combination resulted in a sustained current and only when both peptides were present. Following co-application, low pH currents were potentiated similarly to the effect seen with only PcTx1. This suggests APETx2 and PcTx1 bind rASIC1b at different sites to exert each respective activity. Although there is a possibility that APETx2 binds ASIC3 (inhibitor) and ASIC1b (potentiator) at different sites, it is more likely to be similar to PcTx1 and bind the same region on different ASIC subtypes and produce different activities.

Figure 5.3: Functional competition studies of APETx2 with PcTx1 at rASIC1b. (A) Chicken ASIC1a with PcTx1 bound at the acidic pocket (PDB code 4FZ1). (B) Voltage-clamp recordings of ASIC1b- expressing oocytes showing the effect of APETx2 (black; 10 M) and PcTx1 (pink; 100 nM) alone, co-applied, or co-applied after APETx2 pre-incubation. Scale bar shown in grey (abscissa scale 10 s, ordinate scale 2000 nA). Analysis of the peptide-induced effects for (C) peak current and (D) the co-application induced sustained current at rASIC1b. Data points represent mean ± SEM (n ≥ 4).

Functional competition was also performed at rASIC3 where APETx2 is an inhibitor (IC50

160 nM) and MitTx an agonist (EC50 830 nM). The crystal complex of cASIC1a with MitTx shows binding along the entire thumb domain (Figure 5.4A). Application of APETx2 (1 M) to rASIC3 did not prevent MitTx (1 M) induced channel opening. In fact the rate of opening was increased with APETx2 pre-conditioning as opposed to MitTx application alone (Figure 5.4B and C). In contrast, co-application of APETx2 and MitTx to a channel pre-opened with MitTx resulted in a transient inhibition of current, followed by a reversion to the sustained open state (Figure 5.4B and C). This suggests that pre-application of either peptide does 68 not prevent the activity of the second peptide at ASIC3. The combined data at ASIC1b and ASIC3 suggest that APETx2 does not bind to the same site as either PcTx1 or MitTx, and agrees with the mutagenesis work that APETx2 does not bind to either the acidic pocket or thumb domain.

Figure 5.4: Functional competition studies of APETx2 with MitTx at rASIC3. (A) Chicken ASIC1a with MitTx bound to the thumb domain (PDB code 4NTW). (B) APETx2 (black; 1 M) does not inhibit MitTx (blue; 1 M) evoked currents in ASIC3-expressing oocytes (upper panel). Whereas APETx2 is able to partially and transiently block MitTx-evoked current (lower panel). Scale bar shown in grey (abscissa scale 60 s, ordinate scale 1000 nA). (C) Analysis of the APETx2 and MitTx effects at ASIC3. Data points represent mean ± SEM (n ≥ 4).

5.3.4 ASIC1a:1b chimeras to uncover the APETx2 binding site

Given our knowledge of APETx2 selectivity at rASIC1a (weak inhibition) and rASIC1b (potentiator), a series of chimeric channels in which increasing N-terminal segments of rASIC1b replaced rASIC1a were produced. 10 M APETx2 weakly inhibited rASIC1a chimeras containing up to residue 98 of rASIC1b, but potentiated chimera-166 and wild-type rASIC1b (Figure 5.5A). Therefore the potentiating effect of APETx2 at rASIC1b involves a region between residue 98 and 166 (highlighted purple in Figure 5.5B). This stretch of amino acids corresponds to the finger domain of ASICs, and is a potential novel binding site for ASIC modulators. Importantly, this is also in good agreement with other regions of ASICs that have been eliminated by the ASIC3 mutagenesis and functional competition studies.

69

Figure 5.5: ASIC1a:1b chimeras for APETx2 binding site. (A) Effect of 10 M APETx2 at rASIC1a:1b chimeras. Chimeras are increasing amounts of N-terminal ASIC1b replacing ASIC1a up to the splice site at residue 186 (numbering is amino acid position at rASIC1a). Data points represent mean ± SEM (n ≥ 4). (B) Region equivalent to residues 98 to 166 on the chicken ASIC1a closed state structure highlighted in purple (PDB code 5WKU).

5.3.5 APETx2:rASIC1b pharmacophore

Although APETx2 is analgesic in rodents (63,70,72), its subtype selectivity within ASICs and across other channel types (i.e., hERG and voltage-gated sodium channels) suggests that it needs to be made more selective to improve its value as a research tool and for further development as a potential therapeutic lead. We have previously shown a biphasic analgesic response of APETx2 in an inflammatory pain model in rats (Figure 5.6A) (146). The decrease in analgesia at higher doses of APETx2 was suggested to be due to its potentiation of rASIC1b (another peripheral pain target). Here, the activity of several APETx2 mutants on rASIC1b were tested at 10 M – the estimated EC50 for APETx2 potentiation of rASIC1b and thus the concentration where we are most likely to see a shift in activity (Figure 5.6B). Although there is overlap in key functional residues for both ASIC3 and ASIC1b activity, compared to wild type, Y16A has greater selectivity for potentiating ASIC1b than inhibiting ASIC3, whilst R24A has ~10-fold lower ASIC1b activity than wild-type yet only a ~2-fold decrease in ASIC3 inhibitory activity. Thus R24A has ~5-fold improved relative selectivity for inhibiting ASIC3 over wild type APETx2.

70

Figure 5.6: APETx2 interacts with both rat ASIC3 and ASIC1b. (A) Effect of APETx2 on inflammatory pain in rats (n = 10–15) (Data from panel A is taken from a co-first author manuscript not included as part of this thesis; animal studies were performed by Dr Jia Yu Lee) (146). (B) The left hand plot shows inhibitory activity of 1 M APETx2 and mutants at rASIC3, and the right hand plot displays the potentiating activity of 10 M APETx2 and mutants against rASIC1b (n ≥ 5). Data points represent mean ± SEM.

5.4 DISCUSSION

Despite being discovered ~15 years ago, no data has been reported on either the mechanism of action or binding site of the prototypical ASIC3 inhibitor APETx2. In this work we have provided the first insights into these details. We show that APETx2 likely binds to the closed state of ASIC3, and prevents the extracellular proton-induced conformational changes that lead to channel opening (Figure 5.1A–D). The lack of significant shifts in the pH-dependence curves, in particular the activation curve, is similar to the mechanism of inhibition for the spider venom peptide Hi1a at rASIC1a. This is the first report of an ASIC3 inhibitor with this mechanism of inhibition. The binding affinity of APETx2 is likely greatest for the closed state of ASIC3, which prevents the study of its interaction with other states under most physiological conditions. However, APETx2 was able to partially reverse the MitTx-evoked open state of rASIC3 (Figure 5.4B and C). Although a toxin-gated open state, this suggests APETx2 is able to bind to and have a functional effect on the ASIC3 open and closed states.

The combined results of voltage-dependence of inhibition, mutant and chimeric channels, and functional competition with PcTx1 and MitTx suggest binding of APETx2 to the finger domain of ASICs. This is the first time a potent peptide modulator of ASICs has been shown to bind to the finger domain to exert its effect. Interestingly, the finger domain has previously been implicated in the zinc and extracellular protease modulation of ASICs. Lys133 of the

71 finger is involved in inhibition and potentiation of rat and human ASIC1a, respectively (151). Cys195 plays a role in zinc-mediated inhibition of ASIC1b, and His162 in potentiation of ASIC2a (152,153). Furthermore, trypsin and matripase cleavage at Arg145 of rASIC1a results in diminished currents and altered gating properties (154,155). Although the exact conformational changes that occur during ASIC gating have yet to be clarified, the sensitivity of ASICs to compounds binding to the finger clearly show it to be critical for gating. Crystal structures of ASIC1 reveal an expanded finger and thumb domain in the closed state compared to the MitTx-open and desensitised structures. This is consistent with voltage- clamp fluorometry, cysteine accessibility, and luminescence resonance energy transfer studies which report movement of the finger and thumb domains closer to the ASIC core during proton gating. Binding of APETx2 to the finger domain agrees well with its mechanism of action, whereby the finger domain cannot move in response to extracellular protons and initiate the activation cascade when bound by APETx2. Although many regions of ASIC3 were excluded as the binding site, the findings which directly show the finger to be the region involved in APETx2 activity used the rASIC1a:1b chimeras. It is therefore necessary to confirm that this region is indeed the peptide binding site for the target ASIC3 channel. If the binding site is conserved between ASIC subtypes, understanding the molecular details of the interaction will require further work determining the amino acids involved in binding. This could be delineated either via extensive mutagenesis studies or potentially a high-resolution complex structure.

APETx2 potentiation of the peripheral pain target rASIC1b is a possible cause for the biphasic analgesic response provided by APETx2 (146), making this an undesirable off- target activity. Here we have identified several residues on APETx2 that can be targeted for further mutagenesis to optimise the selectivity for either rat ASIC3 or ASIC1b. A better understanding of the binding site will undoubtedly assist in designing further optimised peptide. Nonetheless, Y16A from this study retains ASIC1b potentiation and has greatly reduced ASIC3 activity. Thus could serve as a pharmacological tool to confirm that the analgesia observed at lower APETx2 concentrations is mediated by ASIC3 inhibition, whereas the loss of antihyperalgesia at higher APETx2 concentrations is due to ASIC1b activity. Given APETx2 is still the most potent ASIC3 inhibitor available, production of a more selective analogue would provide a better tool when studying the role of ASIC3 in physiological and disease states.

72

CHAPTER 6

ASIC1 subtype and species dependent activity of mambalgins: not a potent inhibitor of human ASIC1b

73 6.1 INTRODUCTION

Physiological extracellular pH is well regulated at pH ~7.4, and acidic perturbations of sufficient magnitude are able to activate acid-sensing ion channels (ASICs) expressed throughout the nervous system (4,6). In mammals four genes (ASIC1–4) produce at least six ASIC subunits (ASIC1a, 1b, 2a, 2b, 3, and 4; ‘a’ and ‘b’ denote splice isoforms) that can form functional homo- or heterotrimeric cation channels. In recent years, a large focus has been placed on the involvement of ASIC1b in peripheral pain pathways where acidosis is associated with inflammation and trauma (18,76,88,156). Genetic approaches have helped explain the role of ASICs in physiology and knockout studies uncovered their role in pain pathways. However, many ASIC-knockout studies have yielded differing phenotypes and somewhat contrasting results (13). Strong evidence implicating ASICs in pain pathways has come from the use of pharmacological tools (33). The non-selective ASIC agonist MitTx evokes pain responses when injected into the mouse hindpaw, an effect that is ablated in ASIC1-knockout animals (76). MitTx is a potent agonist of both ASIC1a and 1b subtypes, making it unclear which subtype is more important for peripheral analgesia. Mambalgins are the first reported potent inhibitors of rodent ASIC1b, and their use, in addition to supporting data from siRNA mediated gene silencing, demonstrated that ASIC1b, and not ASIC1a, is important in peripheral pain sensing in mice (18).

Mambalgins are a group of three-finger toxins isolated from black and green mamba snake venoms (Figure 6.1A) (18,31). They inhibit homomeric ASIC1a with varying reported potency and efficacy (IC50 rat: 3–55, human: 127 nM, and chicken: 124 nM), and partially inhibit rat ASIC1b (rASIC1b; IC50 44–192 nM), as well as several heteromers containing rat ASIC1a and 1b (18,89). Mambalgin-2 (Ma-2) was shown to inhibit rASIC1a via decreasing its affinity for protons, which can be seen as an acidic shift of the pH-dependence of activation. A combination of rASIC1a:2a chimeric studies, rASIC1a and chicken ASIC1a (cASIC1a) channel mutagenesis, and a Ma-1:cASIC1a cryo-EM complex at 5.4 Å suggest mambalgins bind predominantly to the thumb domain of ASICs (45,50,89,157). However, these studies propose different peptide binding poses and provide insufficient detail to explain the ASIC species and subtype selectivity of mambalgins. The partial pharmacophore of Ma-1 for rASIC1a was proposed to comprise Phe27, Arg28, Leu32, Ile33, and Leu34 in a study limited to mutants in finger 2 of the peptide (44). Subsequent work showed His6 of finger 1, along with Phe27 and Arg28 to be important for activity at cASIC1a, as well as proposing interactions between Met25, Pro26, Leu30, and Leu32 with the channel (89).

74 Again, these studies proposed vastly different binding poses that yield significantly different mambalgin pharmacophore residues. Development of mambalgins as effective pharmacological tools and clinical candidates will require clarification of the interactions between mambalgins with specific ASIC variants, especially on the human channels. Thus, here we have performed a more extensive mutagenesis and a broader mechanism of action study to reveal the extent of involvement that direct binding interactions and mechanism of action play in determining the ASIC subtype and species specificity of mambalgins.

6.2 METHODS

6.2.1 Peptides

Recombinant mambalgins were produced as described in Chapter 2 (Figure 6.1), synthetic mambalgin-2 (sMa-2) provided by Thomas Durek (University of Queensland, Brisbane, Australia) was produced as previously described (45), and native mambalgin-3 (nMa-3) was isolated by Kirsten Voll (previously University of Queensland; currently Boehringer Ingelheim, Munich, Germany) from Dendroaspis angusticeps using a combination of ion- exchange chromatography and RP-HPLC.

Figure 6.1: Production of recombinant Ma-3. (A) Amino acid sequence of mambalgins with identical residues shown as a dash. Ma-1 and Ma-2 were isolated from the black mamba (Dendroaspis polylepsis polylepsis) and Ma-3 from the green mamba (Dendroaspis augusticeps). N-terminal serine for Ma-3 (bold and italicised) is a vestige from the TEV cleavage site for recombinant material. (B) RP-HPLC purification of TEV cleavage reaction to yield recombinant Ma-3 liberated from the

His6-MBP fusion tag. (C) RP-HPLC chromatogram of pure recombinant Ma-3 showing a single peak.

75 (D) MALDI-TOF MS spectrum in linear mode showing the M + H+ ion for purified recombinant Ma-3 (observed, 6655.05; calculated, 6654.72; average rather than monoisotopic). The recombinant product contains an additional N-terminal serine that is a vestige from the TEV cleavage site.

6.2.2 Data analyses

Changes in IC50 of >3-fold were considered different to wild type. Multiple comparisons were performed using a one-way ANOVA with Dunnett’s test. On- and off-rate kinetics were calculated by fitting the data to a single exponential function.

6.3 RESULTS

6.3.1 Ma-3 subtype selectivity and mechanism of action

The inhibitory activity of recombinant Ma-1 and Ma-3 were tested using the voltage-clamp method on oocytes expressing rASIC1a and showed an IC50 of 4.93 and 2.35 nM, respectively. This activity is comparable to that of synthetic Ma-2 (sMa-2 IC50 3.33 nM) and native Ma-3 (nMa-3 IC50 2.86 nM) at rASIC1a (Figure 6.2A), as well as published values at this channel (18,45). Ma-3 inhibited rASIC1b with an IC50 of 32.6 nM, however the inhibition was incomplete with an Emax of ~65% (Figure 6.2B). Ma-3 potently inhibited hASIC1a with an IC50 of 18.6 nM (Figure 6.2C and D). Surprisingly, the mambalgins have yet to be tested at hASIC1b. We found the activity of Ma-3 at hASIC1b to be drastically different to that at rASIC1a, rASIC1b, and hASIC1a (Figure 6.2C and D). Application of Ma-3 at pH 7.4 and activating hASIC1b with pH 6 resulted in current potentiation with an EC50 of 60.0 nM. In contrast, when using a more intense activation stimulus of pH 5, Ma-3 weakly inhibits hASIC1b (IC50 > 1 M).

76

Figure 6.2: Mambalgin subtype selectivity. (A) Comparison of rASIC1a inhibitory activity of recombinant, synthetic, and native mambalgins. (B) Concentration-response curves for inhibition of homomeric rat ASIC1a and 1b by Ma-3. (C) Example traces and (D) concentration-response curves for Ma-3 activity at human ASIC1a and 1b. All channels were conditioned at pH 7.4, and currents stimulated by pH 6.0 with the exception of hASIC1b that was tested at pH 5 and pH 6 stimulation. Data are mean ± SEM, and n = 5–10.

The mambalgin mechanism of action has only been demonstrated for rASIC1a and the data presented such that the curves in the presence of mambalgin were normalised to their own maximum, rather than the maximum of the control curves thus not demonstrating the effect of the peptide on efficacy (18,158). Here, similar experiments were performed using rat and human ASIC1a and ASIC1b (Figure 6.3A–I). The pH50 of activation of both rASIC1a and hASIC1a was shifted to more acidic values by 3 and 10 nM, respectively (Figure 6.3A and

B). The smaller degree of shift in pH50 here (+0.2 pH units by 3 nM Ma-3) is likely due to a lower concentration of peptide used compared to published values for rASIC1a (pH50: control = 6.39 and 10 nM Ma-2 = 5.70) likely due to a lower concentration of peptide used (158). Ma-3 did not shift the pH-dependence of activation of rASIC1b, however it inhibited the current regardless of the pH used for stimulation (Figure 6.3C). The effect of Ma-3 on the pH-dependence of hASIC1b activation was very different (Figure 6.3D). We noted a shift in the pH50 to more alkaline values (i.e. less protonation required to induce same level of activation), combined with a depression in current compared to control when stimulated with

77 more acidic pH values (pH 5). This mechanism of action explains the apparent dual pharmacology of Ma-3 at hASIC1b when obtaining concentration-response curves using different stimulating pH (Figure 6.2C and D) . As previously described for Ma-2 and rASIC1a

(18,158), Ma-3 had minimal effect on the pH50 of steady-state desensitisation (SSD) of rat and human ASIC1a (Figure 6.3E and F). Ma-3 produced a modest acidic shift (0.12 pH units) in the SSD properties of rASIC1b (Figure 6.3G), whereas a drastic shift in SSD to more alkaline values was observed for hASIC1b (0.29 pH units; Figure 6.3H).

Figure 6.3: Ma-3 mechanism of action. Effect of Ma-3 on the pH dependence of (A–D) activation and (E–H) steady-state desensitisation (SSD) for rASIC1a, hASIC1a, rASIC1b, and hASIC1b without (control) and with 3, 10, 30, and 1000 nM Ma-3, respectively. Black line and circles is control without Ma-3. Solid red line and closed squares is in the presence of Ma-3 with data normalised to maximal control current (I/Icontrol max). Dashed red line and open squares is in the presence of Ma-3 and normalised to the maximal current evoked with Ma-3 (I/IMa-3 max). SSD curves were obtained by activating channels with pH 5 stimulus. (I) pH50 values for data in the presence and absence of Ma- 3 (normalised to control) from panels A–H fit with the Hill equation. Data are mean ± SEM (n = 6–8).

78 The calcium-dependence of rASIC1a inhibition by Ma-3 was not significantly different between in 0.1 and 1.8 mM extracellular calcium, however a ~2.5 fold shift to higher IC50 occurred in the presence of 10 mM calcium (Figure 6.4A). rASIC1a inhibition upon exposure to 30 nM Ma-3 is rapid, with a half-time of 8.44 s and complete activity within 50 s (Figure 6.4B). Recovery from inhibition was also fast when rASIC1a is stimulated by protons immediately after the Ma-3 incubation and every 60 s following (see protocol in Figure 6.4C). However, Ma-3 wash-off is considerably slower when channels are washed out with continuous flow of peptide-free solution (pH 7.45) after Ma-3 incubation, and the first post- peptide proton-stimulus occurs after a given time period (half-time of 92 s; Figure 6.4C).

Figure 6.4: Ma-3 calcium-dependence and kinetics of activity at rat ASIC1a. (A) Calcium- dependence of rASIC1a inhibition by Ma-3. (B) The on-rate for inhibition of rASIC1a by 30 nM Ma-3 (i.e. degree of inhibition after application of peptide for varying periods of time). (C) The off-rate of Ma-3 was assessed after applying 30 nM peptide for 180 s, then stimulating the channel at one minute intervals (filled circles and solid line), or washing with peptide-free control solution and stimulating after varying periods of time (open squares and dashed line). The time constant (t1/2) for the on- and off-rates were calculated by fitting the data to a single exponential function. Data are mean ± SEM, and all n = 6–8.

79 6.3.2 Ma-3 pharmacophore

The activity of a panel of 14 of Ma-3 variants was tested at rASIC1a, rASIC1b, and hASIC1b (see Figure 6.5A for mutated residues mapped to the Ma-2 structure). To expand on previously published work, several residues predicted not to be important for function were also mutated, as they will be useful as negative controls to confirm docking studies. In agreement with published reports for activity at rASIC1a (44), we see a >10-fold loss in activity for Ma-3 R28A, K31A, L32A, I33A, and L34A, in addition to a ~3-fold loss for the K57A mutant (Figure 6.5B).

Figure 6.5: Activity of Ma-3 mutants at rat ASIC1a. (A) Cartoon representation of Ma-2 NMR structure with mutated residues highlighted and depicted as sticks (PDB code 2MFA). (B) Concentration-response curves for inhibition by wild type Ma-3 and mutants of rat ASIC1a. Data are

# mean ± SEM (n = 5–8), and denotes >3-fold change in IC50 from wild type.

80 Mambalgin evoked peripheral analgesia in rodents is through inhibition of the ASIC1b subtype, however pharmacophore details have only been determined at rASIC1a. We tested our panel of Ma-3 mutants at rASIC1b and showed that the R28A, K31A, L32A, I33A, and L34A mutants had the most prominent decrease in activity, and the K57A and K57E mutants lost ~5-fold activity (Figure 6.6).

Figure 6.6: Activity of Ma-3 mutants at rat ASIC1b. Concentration-response curves for inhibition by wild type Ma-3 and mutants of rat ASIC1b. Data are mean ± SEM (n = 5–8), and # denotes >3-fold change in IC50 from wild type.

Given the lower potency of Ma-3 at hASIC1b and pharmacology that is dependent on the stimulating pH, peptides were screened at 1 M for activity on both pH 5 and pH 6 stimulated currents. Interestingly, the potentiation of pH 6 currents by Ma-3 was significantly attenuated with the R28A, L32A, and L34A mutants (Figure 6.7A). In contrast, the inhibition of pH 5 currents by Ma-3 was decreased for the mutants R28A, K31A, L32A, I33A, and L34A (Figure 6.7B), similar to the residues important for inhibition of rASIC1a and 1b.

81

Figure 6.7: Activity of Ma-3 mutants at human ASIC1b. Activity of 1 M peptide at hASIC1b when stimulated with (A) pH 6 and (B) pH 5. Data are mean ± SEM (n = 5–8), and * denotes P < 0.05 in activity from wild type (one-way ANOVA).

6.3.3 Ma-3 binding site

The first insight into the mambalgin binding site at ASICs was shown to be in the vicinity of the acidic pocket, as the Ma-2 IC50 at rASIC1a F350A mutant was >50-fold greater compared to wild type channels (45). To better define the interactions between mambalgins and ASICs in this region and understand ASIC subtype and species specificity, a panel of channel mutations surrounding rASIC1a Phe350 were produced. The mutant channels generated were functional with proton-gated currents with similar pH sensitivity to wild type

(data not shown). Given the difference in activity of Ma-3 at rat and human ASIC1a (IC50

2.35 nM and 18.1 nM, respectively) and rat ASIC1a and 1b (IC50 2.35 nM and 32.6 nM, respectively) we used the sequence differences between these channel variants to guide initial channel mutagenesis. 82 Rat and human ASIC1a share a mechanism of Ma-3 inhibition, yet the IC50 differs by ~8 fold. Comparison of the sequence of these channels reveals three potential surface-exposed amino acid differences that could account for the different potency (Figure 6.8A). The rat channel was mutated to contain the corresponding human residues at these positions (Figure 6.8B). No difference in Ma-3 potency was observed for the A178V and 289 DL- insertion mutants compared to wild type rASIC1a. In contrast, rASIC1a N291K shifted Ma- 3 potency to overlap well with that of hASIC1a, and the reverse mutation in hASIC1a (K291N) recovered ~50% of the loss in activity compared to rASIC1a, providing strong evidence that this residue in the lower thumb is directly involved in peptide interactions.

Figure 6.8: Binding site interactions that determine the potency difference of Ma-3 at rat and human ASIC1a. (A) ASIC residues that are not conserved between rat and human ASIC1a and mutated here are highlighted on a rASIC1a homology model based on the cASIC1 closed state (PDB code 6AVE). (B) Concentration-response curves for Ma-3 inhibition of mutant ASICs. Data are mean ±

# SEM (n = 5–8), and denotes >3-fold change in IC50 from wild type.

ASIC1a and 1b are splice isoforms that differ in the first third of the protein (residues 1–186 in rASIC1a). Rat ASIC1a was mutated to contain the corresponding ASIC1b residues at several of these positions along the thumb:palm domain interface (Figure 6.9A). The rASIC1a mutations Q84E, R175C, and E177G (but not E177Q, a rASIC1a to 2a substitution) resulted in a slight loss in activity, moving the sensitivity towards that of rASIC1b (Figure 6.9B and C). However no individual mutation tested could entirely explain the subtype specificity, or the lack of full inhibition observed at rASIC1b.

83

Figure 6.9: Amino acids that differ between rat ASIC1a and ASIC1b in the Ma-3 binding site. (A) ASIC residues that are not conserved between rat ASIC1a and 1b that were mutated here are highlighted on a rASIC1a homology model based on the cASIC1 closed state (PDB code 6AVE). (B)

Concentration-response curves for Ma-3 inhibition of mutant ASICs. (C) Calculated IC50 values from # panel B. Data are mean ± SEM (n = 5–8), and denotes >3-fold change in IC50 from wild type.

In order to determine the full extent of the mambalgin binding site, rASIC1a mutations that spanned a wider surface exposed area surrounding the acidic pocket were produced. Mutants were selected to include the boundary at the top of the thumb and finger domains (E235A and E342G) through to the bottom of the thumb domain (E362R as well as N291K in Figure 8), as well as several mutations in between (Figure 6.10A) There was a ~6-fold loss of activity on the E362R mutant and much larger losses of activity in the Y316A, F350A, and Y358A mutants, highlighting the importance of the middle and lower thumb domain for Ma-3 activity (Figure 6.10B and C).

84

Figure 6.10: ASIC mutations in the wider Ma-3 binding site. (A) ASIC residues in the Ma-3 binding region that were mutated here are highlighted on a rASIC1a homology model based on the cASIC1 closed state (PDB code 6AVE). (B) Concentration-response curves for Ma-3 inhibition of mutant

# ASICs. (C) Calculated IC50 values from panel B. Data are mean ± SEM (n = 5–8), and denotes >3- fold change in IC50 from wild type.

6.4 DISCUSSION

The discovery of mambalgins generated much excitement, as these peptides were shown to abolish pain in mice through inhibition of ASIC1a in central neurons and ASIC1b in peripheral neurons (18). The ASIC1a-mediated analgesia has been reported in mouse models of inflammatory and neuropathic pain (intrathecal administration) (18,88). Mambalgins were also reported to reduce anxiety-related behaviours associated with arthritis in rats following microinjection into the basolateral amygdala (159). Intraplantar and intravenous mambalgins also evoke an opioid-independent potent analgesia via peripheral ASIC1b in rodent models of acute thermal and inflammatory hyperalgesia (18), neuropathic pain (88), and mechanical allodynia associated with migraine (156). These data revealed for the first time the involvement of the ASIC1b subtype in rodent peripheral pain pathways, and discovered a potential analgesic lead as efficacious as morphine without the issues of tolerance and respiratory distress.

85 With the increasing interest in furthering our understanding of mambalgins and their interaction with ASICs, several groups have developed elegant strategies for its chemical synthesis (44,45,90,160). In this study we used an E. coli periplasmic expression system for the successful production of recombinant mambalgins. The microgram quantities of peptide produced with this system are lower than the synthetic approaches, however do not require oxidative folding making it a time effective option to produce sufficient quantities for in vitro studies. The activity of recombinant wild-type Ma-3 was confirmed on rASIC1a, and the activity screened across rat and human ASIC1a and 1b. Despite the great interest in the peripheral analgesia provided by mambalgin inhibition of rodent ASIC1b, we show for the first time that mambalgins are not potent inhibitors of human ASIC1b. This has grave implications for translating rodent studies to a clinical setting. Interestingly, the reported expression profiles for ASIC1a and 1b also appear to be different between rodents and humans (39). Mambalgin-evoked peripheral analgesia of wild type and ASIC1a-/- mice across multiple rodent models is of no surprise, as ASIC1b is expressed almost exclusively in peripheral nociceptors of rodents and with greater expression levels than ASIC1a (39,161). In contrast, ASIC1a expression levels in human DRG is ~4-fold greater than ASIC1b (39). In humans, transdermal acid-induced pain is alleviated by subcutaneous amiloride suggesting the involvement of ASICs (162,163). However, amiloride is a non- selective ASIC blocker making it unclear which ASIC1 subtype is more important for peripheral pain in humans. If the ASIC subtype expression profile is different between species it begs the question; do humans have the same ASIC-associated nociceptive pathways as rodents? Which ASIC1 isoform is more important for mediating acid-induced pain in humans, and will a human ASIC1b inhibitor be analgesic as observed in rodent models? Despite these questions that need answering, increasing our knowledge of the mambalgin:ASIC interaction will assist in the better use of mambalgins as research tools, and may allow for designing a peptide analogue with the desired pharmacology at specific ASIC1 variants.

The effect of mambalgins on the pH-dependence of activation and steady-state desensitisation at rASIC1a showed that inhibition is primarily due to a shift in the activation curve to more acidic pH (18,158), a mechanism we show is shared for hASIC1a. This is in agreement with the off-rate data at rASIC1a which shows mambalgin unbinding is much faster when exposed to low pH compared to wash-off at neutral pH. This suggests that the binding affinity of mambalgin for rASIC1a is substantially lower at pH 6, and that either surface charge at pH 6 and/or conformational changes that occur to the channel in response 86 to activation are forcing the peptide off the binding site. Inhibition of rASIC1b is via a subtly different mechanism that is pH-independent when applied at physiological pH, thus the degree of inhibition is unchanged regardless of the stimulating pH. Nevertheless in all three channel variants mambalgins bind the closed state and prevents proton-evoked activation gating, resulting in channel inhibition. Human ASIC1b is not potently inhibited by mambalgins, but rather peptide application increases the affinity for protons, as reflected by the alkaline shift in the pH-dependence of both activation and steady-state desensitisation. This mechanism promotes ASIC gating which leads to potentiation of pH gated currents when stimulated with pH ~5.75–6.5, and also drastically increases the pH at which steady- state desensitisation is induced. It could be that with low pH (<5.75) stimulus inhibition is observed, as mambalgins in combination with the increased proton concentration is able to promote desensitisation of channels in a manner similar to PcTx1. State-dependent mambalgin binding is unsurprising as cASIC1a crystal structures show significant conformational changes in the extracellular domain between the closed state and open or desensitised (28). Furthermore, the ASIC1 modulating peptides PcTx1, Hm3a, and Hi1a, which also interact with the thumb domain, show state-dependent binding (43,48,60). PcTx1 binds both rASIC1a and 1b in the acidic pocket, however the functional outcome is different at these variants (37,48). This difference in pharmacology is due to a preference for binding the desensitised state of rASIC1a and the open state of rASIC1b, which appear to have almost identical three-dimensional conformations (25,27). In contrast, mambalgins are able to bind both the ASIC desensitised conformation (hASIC1b activity) and the resting conformation (rASIC1a, 1b, and hASIC1a activity), which are structurally different in the mambalgin binding site. This is the first report of an ASIC gating-modifier peptide where the variants-dependent mechanism of action and activity is due to this preference for totally different conformational states of a given channel variant.

Mutagenesis studies of the interactions between mambalgin and ASICs have largely been centred on ASIC1a. The mambalgin pharmacophore for activity at rASIC1a has been studied, however, that work was limited to alanine mutants in and around finger II of synthetic Ma-1 (44). Mourier et al. performed full concentration-response curves, and showed >10- fold loss in IC50 for mutants F27A, R28A, L32A, I33A, and L34A, concluding that these residues are crucial for the ASIC1a interaction. A second study used cASIC1a and tested the inhibitory activity of several Ma-1 mutants (Q5A, H6A, K8A, F27A, and R28A) at a single concentration of 500 nM, extending the peptide mutagenesis to include finger I (89). Using a single concentration of Ma-1, it is difficult to determine the true extent of loss in activity of 87 each mutant peptide. For example both a 10- and 1000-fold shift in IC50 would be equally inactive when tested at 500 nM. Nonetheless, this study reported a statistically significant loss in activity for Ma-1 mutants H6A, F27A, and R28A, which suggests mambalgin finger I is also important for its effect at ASIC1a.

Here, we have expanded the mambalgin structure-activity relationship by mutating residues outside fingers I and II, as well as studying the activity of several mutants at rat and human ASIC1b along with rASIC1a. The mutants tested here agree with the published data at rat and chicken ASIC1a that the most important residues for activity at ASIC1a are in finger II. We also confirm that residues that are spatially distant (Arg14, Ser39, Ser40, and Arg55) from the key pharmacophore residues are not involved in the mambalgin interaction with rASIC1a. Residues Arg28, Lys31, Leu32, Ile33, and Leu34 were also critical in the interaction with rASIC1b, as well as a smaller loss in activity for K57A and K57E. Although these subtype variants show slightly different activity and mechanism of action to wild-type mambalgins, it appears the binding interface is in large part shared between channels. Given the lower potency of mambalgin for hASIC1b we studied the activity at a single concentration, however face the same limitation as the previously described experiments at cASIC1a. Surprisingly, the potentiating activity with pH 6 stimulus was retained in most mutants, with a significant loss observed for only some finger II mutants (R28A, L32A, and L34A). With pH 5 stimulus at hASIC1b, finger II mutants R28A, K31A, L32A, I33A, and L34A almost completely lost inhibitory activity, with smaller losses for R14A, R55A, and K57E. There are two major points to consider from the hASIC1b experiments. Firstly, why are Lys31 or Ile33 not important for the potentiation of pH 6 evoked currents? We suspect this could be linked with the mechanism, where in wild type mambalgins, the Arg28 side chain acts in a similar manner to protons to potentiate the pH 6 current. In this case, Arg28 is important to produce the functional outcome, as opposed to Lys31 and Ile33 that are more important for the binding affinity. The two leucine residues make contacts that are important for the actual binding to all channels tested. Secondly, why do more mambalgin residues appear to be important for inhibition of pH 5 evoked hASIC1b currents compared to other conditions tested? The larger interacting surface for hASIC1b inhibition could be an indication that the binding pose for mambalgin at hASIC1b is significantly different to other ASIC subtypes, in agreement with the different activity and mechanism of action we revealed.

88 The binding site of mambalgins on ASICs has been studied by several groups using a combination of functional and structural techniques as well as docking studies. However, there is disagreement between these studies over the proposed binding orientation of mambalgins at ASICs (see Figure 6.11 for published models) (44,45,50,89,157).

Figure 6.11: Comparison of published binding models of mambalgin to ASIC. (A) First docking model of mambalgin-2 on a homology model of rat ASIC1a, using blind docking with ZDOCK (version2.3.2f). Figure 1A and B from (50). This research was originally published in the Journal of Biological Chemistry. Miguel Salinas, Thomas Besson, Quentin Delettre, Sylvie Diochot, Sonia Boulakirba, Dominique Douguet, and Eric Lingueglia. Title. Binding Site and Inhibitory Mechanism of the Mambalgin-2 Pain-relieving Peptide on Acid-sensing Ion Channel 1a J. Biol. Chem. 2014; 289: 13363-13373. © the American Society for Biochemistry and Molecular Biology. (B) Second model of mambalgin-1 on a model of rat ASIC1a by the same group as in Panel A. ZDOCK (server version 2.3.2f) predictions were filtered out to include the pair-wise interaction of rASIC1a-Phe350 with mambalgin-1-Leu32 that was experimentally determined. Figure 9 from (44). Location. This research was originally published in the Journal of Biological Chemistry. Gilles Mourier, Miguel Salinas, Pascal Kessler, Enrico A. Stura, Mathieu LeBlanc, Livia Tepshi, Thomas Besson, Sylvie Diochot, Anne Baron, Dominique Douguet, Eric Lingueglia, and Denis Servent. Title. Mambalgin-1 Pain-relieving Peptide, Stepwise Solid-phase Synthesis, Crystal Structure, and Functional Domain for Acid-sensing Ion Channel 1a Inhibition J. Biol. Chem. 2015; 291: 2616-2629. © the American Society for 89 Biochemistry and Molecular Biology. (C) 5.4 Å cryo-EM complex of mambalgin-1 bound to cASIC1a produced by rigid body fitting the cASIC1a desensitised crystal structure (PDB code 4FZ1) extracellular domain and mambalgin-1 crystal structure (PDB code 5DU1) into the cryo-EM density, followed by further refinement using COOT to better fit density. Figure 3A and B from (89). Location. This research was originally published in Cell Discovery 4, Article number: 27 (2018) at https://www.nature.com/articles/s41421-018-0026-1#Abs1. Title. Cryo-EM structure of the ASIC1a– mambalgin-1 complex reveals that the peptide toxin mambalgin-1 inhibits acid-sensing ion channels through an unusual allosteric effect. Authors. Demeng Sun, You Yu, Xiaobin Xue, Man Pan, Ming Wen, Siyu Li, Qian Qu, Xiaorun Li, Longhua Zhang, Xueming Li, Lei Liu, Maojun Yang, and Changlin Tian. Licensed under CC BY 4.0, with full terms at https://creativecommons.org/licenses/by/4.0/.

The first report of the mambalgin binding site showed that the IC50 of synthetic Ma-2 was >50-fold lower at rASIC1a F350A mutant over wild-type (45), in good agreement with what we have shown here for Ma-3. The first docking model proposed for mambalgin binding to rASIC1a displayed the peptide as lying flat along the surface of the channel to interact with the thumb, beta-ball and palm domains (the acidic pocket; Figure 6.11A) (50). As much of this work used chimeric channels and a single concentration, it is extremely difficult to tease apart the extent to which binding interactions or altered chimeric channel characteristics played a role in the disrupted mambalgin activity. A subsequent study from this group proposed a binding mode significantly different to their first model, with the peptide finger II binding deep into the acidic pocket (Figure 6.11B) (44). This model was refined with functional data (mutant-cycle analysis) suggesting a strong proximity between mambalgin Leu32 and rASIC1a Phe350 residues. The most recent study reported a 5.4 Å cryo-EM structure of cASIC1a in complex with Ma-1, and suggested mambalgin bound exclusively to the thumb domain (Figure 6.11C) (89). This novel binding orientation was supported by mutagenesis of cASIC1a thumb domain residues. The cASIC1a mutants Y317A, D346A, F351A, and D356A were potentiated (all >200% of control currents) by 500 nM Ma-1 in contrast to the inhibition of wild type cASIC1a. Due to the low local resolution of the cryo- EM structure, de novo structure determination was not possible and the authors rigid body fit the density of the cASIC1a (PDB code 4FZ1) and Ma-1 (PDB code 5DU1) crystal structures into the cryo-EM density map. The greatest limitation of this approach, as well as for previous docking studies, is that the ASIC1a structures used for model building were either in the open or desensitised states. It is now known that the thumb domain undergoes rearrangement between the closed and open/desensitised states (28), and here we show that mambalgins inhibit via binding to and stabilising the closed state of rat and human

90 ASIC1a. The ASIC1a binding surface that mambalgins are exposed to in the closed state is most likely significantly different to what has been modelled in studies to date.

Here we focussed our efforts on studying individual point mutations in rASIC1a and performed full concentration-response curves. The 10-fold shift in ASIC1a species- dependent potency between rat and human isoforms can be explained almost entirely by a single point mutation at residue 291 (Asn in rat and Lys in human). This is in stark contrast to the 10-fold difference in rat vs human ASIC1a potency for PcTx1, which we showed is entirely dependent on the inherent channel characteristics and mechanism of inhibition for this peptide (see Chapter 4). This agrees with our mechanism of action studies showing that the mambalgin effect at rat and human ASIC1a species variants is predominantly due to an acidic shift in the pH-dependence of activation. In examining the difference between rat ASIC1a and 1b activity, we show that three mutations along the palm domain (Q84E, R175C, and E177G) slightly lost inhibitory activity. No single mutation could recapitulate the ~15 fold difference in potency between these splice isoforms, or the incomplete inhibition at wild type rASIC1b. Although a triple mutant or further residues may explain the different pharmacology between these isoforms, we predict that the different functional outcome is due to a combination of binding interactions and mechanism of action. In studying further residues throughout the thumb domain that can eventually be used for guided docking studies, we reveal Tyr316, Phe350, Tyr358, and Glu362 to be important for mambalgin activity at rASIC1a. Together these mutants show that the middle and lower thumb domain are the regions making the most important interactions for mambalgin activity.

A global picture of mutagenesis results here and previously published data for both mambalgins and ASIC1a suggests the peptide binds flat along the thumb domain. Figure 6.12 summarises important (red) and not important (green) residues for the interaction between these molecules. The available mambalgin mutagenesis clearly demonstrates the importance of finger II for the ASIC1a interaction, however the data for residues in fingers I and III are either ambiguous or not well studied. This makes it difficult to model the interaction, as several binding poses remain possible. The interacting surface of ASIC1a is better studied, and in contrast to earlier reports suggesting multiple interactions of mambalgins with acidic pocket residues (see Figure 6.11A and B) (44,50), there appears to be only minor interactions in this region. Mutations along the ASIC1a subunit interface (palm domain; point mutants highlighted orange in Figure 6.12) result in slight losses of mambalgin activity, and mutations to the upper and outer thumb do not change the mambalgin effect.

91 The thumb domain is post-splice site for ASIC1a and1b, and thus identical in sequence between these isoforms. Given the binding site is indeed restricted almost entirely to the thumb, this limits the possibility of engineering more ASIC1b selective analogues.

Figure 6.12: Summary of published findings and data presented here of the Ma-3 interaction with ASIC1a. (A) Mambalgin pharmacophore highlighted on the Ma-2 NMR structure (PDB code: 2MFA). Location of the important amino acids and three fingers of mambalgins are shown. (B) The middle and lower thumb domain are involved in the mambalgin interaction with ASIC1a, and amino acids shown to be important mapped onto the extracellular domain of a rASIC1a homology model based on the cASIC1 closed state (PDB code 6AVE). Red residues are critical for the Ma-3:ASIC1a interaction, green residues have been shown not to be important for activity, and orange residues for panel B are Gln84, Arg175 and Glu177 that show a small shift in activity (see Figure 6.9).

92

CHAPTER 7

Discovery and molecular interaction studies of a highly stable, tarantula peptide modulator of acid-sensing ion channel 1

93 Preface to Chapter 7. This chapter is in large part a reproduction of a published manuscript. The peptide (Hm3a) was originally sequenced and identified by Pierre Escoubas in a screen for ASIC1a modulators as described in Section 7.3. Several years later this same venom and peptide was identified in a screen I performed for my thesis. Combining these two independent findings I conceptualised the below project with Lachlan D Rash. Experimental contributions are as follows. Identification and purification of Hm3a: Pierre Escoubas, Ben Cristofori-Armstrong, and Sing Yan Er. Recombinant expression: Ben Cristofori-Armstrong. Subtype selectivity: Ben Cristofori-Armstrong and Sing Yan Er. Mechanism of Action: Ben Cristofori-Armstrong. Channel binding site: Ben Cristofori-Armstrong and Sing Yan Er. Stability studies Sing Yan Er and Ben Cristofori-Armstrong. Drafting of manuscript: Ben Cristofori-Armstrong and Sing Yan Er. Final figures: Ben Cristofori-Armstrong. Editing: Lachlan Rash and Ben Cristofori-Armstrong.

7.1 INTRODUCTION

Over the last two decades, the growing knowledge on the molecular basis of various channelopathies and neurological disorders has fuelled an intense interest in the discovery and development of ion channel modulators as therapeutics and pharmacological tools (164). Spiders, in particular, have proved to be a rich source of such modulators. Their venoms contain numerous biologically active peptides that potently target a variety of ion channels (165,166). As such, peptides isolated from spider venoms have been widely used to dissect the molecular mechanisms of channel function and their role in biological processes (167-169). In this respect, the acid-sensing ion channels (ASICs) are no different with a spider venom peptide being the first, and still the most potent, selective modulator known for ASICs.

ASICs are members of the degenerin/epithelial sodium channel family that are widely distributed in the nervous system and many non-neuronal cells (170). They are primary mammalian acid sensors and as such are activated by protonation (4). Six ASIC subtypes are known (ASIC1a/b, ASIC2a/b, ASIC3 and ASIC4), which can assemble as functional homotrimeric (ASIC1a/b, ASIC2a and ASIC3 only) or heterotrimeric channels (19,22). Each subunit consists of a large extracellular domain, two transmembrane helices, and short intracellular N- and C-termini. The combination of subunits in each channel affects the pH sensitivity, pharmacology, and kinetics of activation and desensitization (22). ASIC1a is the most abundant ASIC subunit in the mammalian central nervous system. Unlike other ASICs, ASIC1a is also permeable to Ca2+, a property that has been associated with ASIC1a’s role 94 in acidosis-induced neuronal injury (56,171). ASICs have been implicated with a variety of pathophysiological conditions such as inflammatory and neuropathic pain (70,76,172), psychiatric illness (7), multiple sclerosis (173), epileptic seizure termination (174) and ischemic stroke (56,57), hence are considered promising therapeutic targets (9).

Although several small molecules have been demonstrated to modulate ASIC activity, they lack subtype specificity and potency, and are limited in their usefulness as pharmacological tools. In contrast, venom peptides are often extremely potent and can have very high selectivity for neuronal targets (165). To date, six ligands targeting various ASIC subtypes have been isolated from animal venoms (32), The prototypical and therefore most studied ASIC ligand is the spider venom peptide, PcTx1, from the spider Psalmopoeus cambridgei

(34). PcTx1 inhibits homomeric ASIC1a with an IC50 of ~0.9 nM, and stabilizes the open state of ASIC1b. More recently it has been shown to also inhibit heteromeric ASIC1a/2b and ASIC1a/2a channels (35,36). The only other inhibitory peptide ligands of ASIC1 include the mambalgins, from the Mamba snakes (Dendroaspis spp.), which have been shown to inhibit

ASIC1a and ASIC1b containing channels with IC50 values between 11 and 300 nM (18,32).

In a functional screening of a panel of spider venoms, we observed ASIC1a inhibitory activity from the venom of the Togo starburst tarantula (Heteroscodra maculata). Here, we describe the isolation, recombinant expression and characterization of a second potent inhibitor of ASIC1a. The peptide named π-theraphotoxin-Hm3a (hereafter Hm3a), was pharmacologically characterised to determine its ASIC subtype-selectivity, binding site, mechanism of action and stability.

7.2 MATERIALS AND METHODS

7.2.1 Venom peptide purification and characterisation

Heteroscodra maculata venom was purchased from Spider Pharm (Yarnell, AZ, USA). The lyophilized venom was dissolved in water (1:10 dilution from original venom volume) and stored at –20 °C until required. A total of 50 μL of the diluted venom was fractionated using RP-HPLC on a C18 column (Phenomenex Kinetex, 250 x 4.6 mm, 100Å) using a gradient of solvent A (0.05% trifluoroacetic acid (TFA) in H2O) and solvent B (90% acetonitrile,

0.043% TFA in H2O) as follows: 5% solvent B for 5 min, 5–20% solvent B for 10 min, 20– 50% solvent B for 30 min, 50–80% solvent for 5 min, at a flow rate of 1 mL/min. Fractions were collected by monitoring eluent absorption at 214 and 280 nm and lyophilized. The dried fractions were dissolved in water and aliquots of each fraction were assayed for activity 95 against rASIC1a expressed in Xenopus oocytes. A second RP-HPLC purification was carried out on the active fraction using a C18 column (Thermo Aquasil, 50 × 2.1 mm, 190Å) and a gradient of 10–30% solvent B over 20 min at a flow rate of 0.3 ml/min. All solvents used were HPLC grade.

7.2.2 Peptide sequencing

The pure native peptide was reduced with 2-mercaptoethanol and alkylated by 4-vinyl- pyridinethen sequenced via N-terminal Edman degradation (477A, Applied Biosystems, Foster City, CA, USA). Sequence homologies were determined using sequences obtained from a search of non-redundant protein databases via the BLAST server. Sequence alignments and percentages of similarity were calculated using BLASTP (175).

7.2.3 Data analyses

Half-life (t1/2) was obtained through a non-linear fit to the data of a “dissociation – one phase exponential decay” equation. Results are reported, in the text or on the figures, as means ± SEM. Statistical analysis and comparison between two groups was performed using Student’s t-test with Welch’s correction where standard deviations were not assumed to be equal. Comparisons of three or more groups were performed with one-way ANOVA and

Tukey’s Test for IC50 comparisons, and a two-way ANOVA and Tukey’s Test for stability assays.

7.2.4 Stability assays

Thermal stability was assessed by dissolving Hm3a, PcTx1 or oxytocin in 10 mM phosphate- buffered saline pH 7.4 to a final concentration of 25 M. Samples were incubated at 55 °C and triplicate samples were taken at t = 0, 4, 8, 24, and 48, quenched with 5% TFA solution, centrifuged at 17 000 g for 25 minutes and analysed using RP-HPLC. The peak corresponding to intact peptide was identified by comparison to an untreated sample. Levels of peptide were quantified by integration of the peak area recorded at 214 nm, and plotted as a percentage of the peak area at t = 0 for each respective sample.

Serum stability was performed using male AB human serum (Sigma-Aldrich). Serum was pre-warmed to 37 °C for 15 minutes and added to lyophilized Hm3a, PcTx1, or oxytocin to a final concentration of 50 M. Triplicate samples were taken at t = 0, 1, 2, 4, 8, 12, 24, and 48h for Hm3a and oxytocin, and t = 0, 2, 4, 24, and 48 h for PcTx1, quenched with 20% TFA

96 solution, and stored at 4 °C for 15 minutes to precipitate serum proteins. Samples were centrifuged at 17 000 g for 25 minutes and analysed using RP-HPLC. The peak corresponding to intact peptide was identified by comparison to an untreated sample and levels of peptide were quantified as described above.

7.2.5 Deposition of protein sequence information

Peptide sequence information derived for Hm3a has been submitted to the publicly accessible UniprotKB database (http://www.uniprot.org/) (accession number C0HKD0).

7.3. RESULTS AND DISCUSSION

Despite sustained interest in ASIC1 as a promising drug target (9), the current array of ASIC1 modulatory ligands available as selective pharmacological tools and therapeutic leads is extremely limited. The difficulty in developing more selective small molecule ASIC inhibitors as therapeutic leads is illustrated by a series of medicinal chemistry studies by the Merck Research Laboratories. Three studies using different scaffolds all resulted in more potent molecules against ASICs, however, all had significant off-target effects resulting in substantial sedation in animal studies (176,177). Based on the discovery of PcTx1 in a spider venom (34) as well as preliminary screening data from past studies (Escoubas, unpublished) we hypothesized that spider venoms, and especially tarantula venoms, should be a valuable source of additional novel ASIC modulators. Discovery of novel ligands, and understanding the molecular basis for differences in their biological properties (such as structural and chemical stability) and functions provides valuable insight to develop improved ligands.

7.3.1 Identification and purification of Hm3a

Screening a panel of spider venoms on rat ASIC1a expressed in Xenopus laevis oocytes revealed that Heteroscodra maculata (1/1000-fold dilution) venom robustly (>90%) inhibits this channel. Assay-guided fractionation led to the identification of a minor, early-eluting peptide (Fraction 12) responsible for the rASIC1a inhibitory activity (Figure 7.1A). The peptide has an observed monoisotopic M+H+ of 4285.05 (Figure 7.1B) and was named - TRTX-Hm3a, hereafter Hm3a, according to a rational nomenclature system for naming peptides from animal venoms (178).

97 The full-length amino acid sequence of Hm3a was determined using Edman degradation and verified using mass spectrometry. The theoretical oxidized monoisotopic m/z (M+H+ = 4285.03) of the Hm3a sequence matched the observed monoisotopic mass, suggesting the presence of a free carboxylic acid at the C-terminus and three disulfide bonds. The Hm3a sequence matched that of another ASIC-inhibiting peptide previously isolated from this venom and named Hm12 (Escoubas, unpublished).

Hm3a shares 82% identity with PcTx1, the prototypical ASIC1a inhibitor from Psalmopoeus cambridgei venom (Figure 7.1C). Hm3a differs from PcTx1 by five point mutations and a three-residue C-terminal truncation. Native Hm3a is highly potent on rASIC1a, inhibiting acid-evoked currents with an IC50 of 2.6 ± 1.3 nM when applied at pH 7.45 (Figure 7.1D). Due to the high sequence identity, conserved cysteine framework, and similar pharmacology between Hm3a and the inhibitor cysteine knot (ICK) fold peptide PcTx1 (PDB code 2KNI), it is highly likely that Hm3a is also an ICK fold (Figure 7.1E).

Figure 7.1: Isolation of Hm3a from H. maculata venom. (A) RP-HPLC chromatogram of crude H. maculata venom indicating the Hm3a containing Frac 12*. Inset: effect of Frac 12* on rASIC1a. (B) Chromatogram showing the final purification step of native Hm3a and MALDI-TOF mass spectrum showing a monoisotopic M+H+ of 4285.05. (C) Amino acid sequence of Hm3a aligned with PcTx1 98 (grey background = conserved residues, red = cysteine residues and disulfide connectivity shown below). (D) Concentration-effect curve for inhibition of rASIC1a by native Hm3a conditioned at pH

7.45 yielded an IC50 of 2.6 ± 1.3 nM (n = 6; error bars denote SEM). (E) Homology model of Hm3a (based on PcTx1 NMR structure, PDB code 2KNI) in blue. PcTx1 NMR structure in green, with the disulfide bridges in both shown in red, and N- and C-termini are as indicated.

7.3.2 Recombinant production of Hm3a

In order to obtain sufficient material to carry out further pharmacological characterization of Hm3a, we used an Escherichia coli recombinant expression system that we have previously employed to produce many disulfide rich venom peptides including PcTx1 (41,102). Following IPTG induction, the Hm3a-fusion protein was the dominant cellular protein present (Figure 7.2A) yielding ~1 mg of correctly folded peptide per litre of bacterial culture at >95% purity (determined using RP-HPLC and MS) (Figure 7.2B). The observed M+H+ (4372.58) was in accordance with the calculated value (4372.06) for Hm3a with an additional N- terminal serine left over from the TEV recognition site (Figure 7.2B inset). The recombinant variant was subsequently used for all further characterization studies.

Figure 7.2: Production of recombinant Hm3a (A) SDS-PAGE gel showing the expression, purification and TEV cleavage of the Hm3a-MBP fusion protein. Lanes are as follows: MW marker; Uninduced and Induced (IPTG) whole-cell extract; Insoluble and soluble fraction after cell lysis; Flow- through, soluble cell lysate after passed through Ni-NTA; Wash, eluate from 15 mM imidazole wash; Elution, fusion protein eluted by 250 mM imidazole; Concentrated fusion protein before and after cleavage with TEV protease. (B) RP-HPLC chromatogram of purified recombinant Hm3a. Inset is a MALDI-TOF mass spectrum showing the monoisotopic M+H+ of 4372.58 (calculated 4372.06). Recombinant Hm3a contains a vestigial N-terminal serine from the TEV cleavage.

99 7.3.3 Effects of Hm3a on ASICs

Recombinant Hm3a was similarly potent to the native peptide with an IC50 of 1.3 ± 0.2 nM (Figure 7.3A) confirming that the determined sequence is responsible for the observed ASIC1a activity. Furthermore, it shows that the addition of an extra residue on the N- terminus (the serine from the TEV cleavage site) has little effect on the inhibitory activity of the peptide, consistent with the well-tolerated addition of a serine or tyrosine to the N- terminus of PcTx1 (41,53). Having ascertained the functional integrity of the recombinant Hm3a, the remaining characterization studies were performed using the recombinant peptide. The potency and selectivity of Hm3a was assessed using TEVC electrophysiology at a conditioning pH of 7.45 on oocytes expressing homomeric rASIC1a, rASIC1b, rASIC2a, rASIC3, human ASIC1a, hASIC1b, and co-expressed rASIC1a/ASIC1b (a potential mixture of homo- and heteromeric channels) (Figure 7.3A–B). Hm3a is highly selective for ASIC1, with no observable effect on rASIC2a or rASIC3 up to 10 μM (Figure 3A). In stark contrast,

Hm3a strongly potentiated rASIC1b (~4-fold increase in current amplitude) with an EC50 of 46.5 ± 6.2 nM (Figure 7.3B), as has previously been reported for PcTx1 (37). The potentiating effect was also observed on heteromeric rASIC1a/ASIC1b channels with higher potency (EC50 of 17.4 ± 0.5 nM), but lower efficacy (less than 2-fold increase in current amplitude) as compared to homomeric rASIC1b (Figure 7.3B).

We next assessed species-dependent effects of Hm3a on ASIC1. Similar to the rat isoforms,

Hm3a also inhibited hASIC1a (IC50 of 39.7 ± 1.1 nM) and potentiated hASIC1b (EC50 of 178.1 ± 1.3 nM); it was however ~30-fold, and ~3.8-fold less potent, respectively, than on the rat isoforms (Figure 7.3C). The Hill coefficients from the concentration-effect curves of Hm3a on both rat and human ASIC1a and ASIC1b were 1.3–1.6, suggesting positively cooperative binding of more than one Hm3a peptide per ASIC1 trimer. This is consistent with the co-crystal structures of cASIC1 and PcTx1 that showed up to three PcTx1 peptides bound per channel (see Figure 7.5A) (27,52).

100

Figure 7.3: Concentration-effect curves of recombinant Hm3a for (A) homomeric rASIC1a, rASIC2a, and rASIC3 (n = 6–10), (B) homomeric rASIC1b and heteromeric rASIC1a/ASIC1b channels (heteromeric rASIC1a/ASIC1b is the response observed in oocytes co-injected with both rASIC1a and rASIC1b cRNA) (n = 6), and (C) homomeric hASIC1a and hASIC1b (n = 7). (D) Concentration- effect curves of Hm3a_WT and Hm3a_P38 mutant on homomeric rASIC1a (n = 4–10). *Statistical comparison of logEC50 between rASIC1b and heteromeric rASIC1a/ASIC1b data sets (panel B); p =

0.0052 and F = 8.069, and logIC50 for Hm3a WT and P38 mutant data sets (panel D); p = 0.0029 and F = 9.388. Conditioning pH used to generate all the above data was pH 7.45, and example traces are shown in Figure S7.1. Data are mean ± SEM.

In the study by Chen et al. (37) that first demonstrated the potentiation of ASIC1b by PcTx1, the authors were unable to determine an accurate EC50 as a plateau in the PcTx1-induced potentiation effect was not reached. Here, we were able to show a maximal response for

Hm3a potentiation of both rat and human ASIC1b allowing more robust comparison of EC50 values. PcTx1 has been shown to inhibit both ASIC1a homomers (34) as well as rASIC1a/ASIC2b and rASIC1a/ASIC2a heteromers (35,36), however, no activity has been reported on channels resulting from rASIC1a/ASIC1b co-expression (presumably containing heteromers). Our results indicate that rASIC1a/ASIC1b heteromers are sensitive to Hm3a in the mid-nanomolar range. A better understanding of the interaction of these peptides with both ASIC1b and ASIC1a/1b heteromers is crucial when they are used as ASIC1a inhibitors in animal studies because both ASIC1a and ASIC1b are highly expressed in the peripheral 101 nervous system in rodents (23). The use of high concentrations of Hm3a or PcTx1 may lead to confounding results, and possible undesirable side-effects, because inhibition of ASIC1b has caused analgesic effects in inflammatory and nociceptive pain models in ASIC1a- knockout mice (18).

Despite the high sequence identity of Hm3a to PcTx1, Hm3a is slightly less potent on rASIC1a (~3-fold compared to PcTx1 in our previous studies) (41,51) (Figure 7.1D and 3A). From our previous structure-activity studies on PcTx1, we have shown that truncation of the three C-terminal residues (Pro38, Lys39, Thr40; variant named Δ3C) resulted in a ~4-fold decrease in activity (51). Furthermore, a P38A mutation of PcTx1 was also ~4-fold less potent than wild-type PcTx1, leading us to the conclusion that the channel interaction mediated by Pro38 in PcTx1 could account for the slight loss in activity observed in Δ3C (51). Being equivalent in length to the PcTx1_Δ3C variant, Hm3a also lacks a proline at position 38. Hence, we hypothesized that the potency of Hm3a on rASIC1a could be enhanced by the addition of a proline at the C-terminus (Hm3a_P38). Hm3a_P38 had an

IC50 of 0.4 ± 0.1 nM (Figure 7.3D) or a 3.3-fold increase in potency as compared to wild-type recombinant Hm3a, in complete agreement with our prediction.

Extensive mutagenesis of PcTx1 (41,51), along with two co-crystal structures in complex with chicken ASIC1 (27,52) have been instrumental in defining the PcTx1 pharmacophore. Three of the residue differences between PcTx1 and Hm3a (D2P, G9S, and E19A) were not identified as potential contacts in structural studies. As a result, the effect of these mutations in PcTx1 have not been studied, and here we show that these mutations do not appear to be crucial for activity against rASIC1a. Thr37 of PcTx1 was identified as a crystal contact, however a PcTx1_T37A mutant was shown to be equipotent with wild-type PcTx1 (51), in good agreement with the Hm3a potency at rASIC1a (Thr37 is a valine in Hm3a). The only residue difference between Hm3a and PcTx1 that falls within the key pharmacophore residues is R28K. Arg28 in PcTx1 makes multiple channel contacts and the PcTx1_R28A mutant had 14.5-fold lower activity on rASIC1a (51). In the case of Hm3a, the conservative mutation of arginine to lysine at position 28 appears to have little to no deleterious effect on peptide activity, thus the substitution is both functionally and structurally neutral. Although lysine has one less amino group, full retention of activity suggests that not all the interactions proposed in crystal studies are making energetically important contributions.

102 7.3.4 Mechanism of action of Hm3a on ASIC1

The homologous peptide PcTx1 mimics the binding of protons in the acidic pocket of

ASIC1a, and inhibits channel currents by shifting the pH of half-maximal effect (pH50) of steady-state desensitization (SSD) and activation to more alkaline values, pushing channels into a desensitized (non-conducting) state at pH 7.45 (48). Thus we assessed the effects of 30 nM Hm3a on SSD of rASIC1a (Figure 7.4A). Application of Hm3a in the conditioning solution induced a parallel alkaline shift in the pH50 SSD of rASIC1a by ~0.37 pH units from 7.24 ± 0.02 to 7.61 ± 0.02 (Figure 7.4C). This shift in SSD by Hm3a pushes ASIC1a into a desensitised state at pH 7.45 (i.e. inhibition), however at 30 nM the magnitude of the shift is insufficient to induce inhibition when applied at conditioning pH 7.9. Therefore at pH 7.9, the channels are in the resting state but in the presence of Hm3a. This allows us to study the effect of the peptide on the pH-dependence of channel activation, without the peptide inducing channel inhibition (Figure 7.4B). The pH50 of activation (pH50 ACT) of rASIC1a under control conditions was 6.08 ± 0.04, as compared to 6.28 ± 0.06 when conditioned with 30 nM Hm3a (Figure 7.4C). Thus Hm3a lowered the proton concentrations required for activation by ~0.20 pH units. The alkaline shift caused by Hm3a in the pH of both SSD and activation confirms the analogous mechanism of inhibition to PcTx1.

The on-rate of Hm3a on rASIC1a was examined by application of 30 nM peptide at pH 7.45 for increasing periods of time followed by activation at pH 6. Fitting a single exponential function to the time course of inhibition data revealed a time to 50% inhibition (t1/2) of ~16 s, with maximal inhibition reached after ~90 s (Figure 7.4D and S7.2). The slow onset of Hm3a inhibition may not be reflective of the actual binding on-rate for the peptide-channel interaction. Hm3a causes inhibition via inducing SSD of ASIC1a, which in itself is a slowly induced process (plateauing at >100 s for pH 7.05 exposure) (24), and likely takes longer than the actual on rate of Hm3a. In agreement with this, application of 30 nM Hm3a for less than 10 s led to potentiation of currents (Figure S7.3), indicative of a rapid interaction between ASIC1a and Hm3a, similar to previously reported observations with PcTx1 on rASIC1a (48). The on rate of Hm3a was similar to PcTx1 as previously reported using oocyte electrophysiology. Full inhibition of rASIC1a induced by 30 nM PcTx1 was achieved ~150 s after peptide application, with a time constant (Tau) of 52 s (37). Recovery of rASIC1a current after inhibition induced by a 180 s exposure to 30 nM Hm3a at pH 7.45 was complete after ~360 s of peptide washout, with a t1/2 of ~132 s (Figure 7.4E and S7.2). ASIC1a recovers from the desensitized state over a time course of several seconds (time constant

103 ~10 s) (24), hence the slower recovery from Hm3a inhibition can be largely attributed to unbinding of the peptide from the channel rather than simply recovery from SSD. Taken together, the mechanism and kinetics data shows that the homologous peptides Hm3a and PcTx1 have an almost identical effect on ASIC1a.

In order to determine the mechanism of ASIC1b potentiation by Hm3a, we analyzed the decay time constants of whole cell current desensitization. Control ASIC1b currents elicited by a pH drop to 5.0 decayed with a time constant (τ) of 1.28 ± 0.06 s (Figure 7.4F). Hm3a caused a concentration-dependent increase in the time constant of desensitization, and could be fit with the Hill function to yield an EC50 of 35.9 ± 17.2 nM, which is consistent with the EC50 determined for potentiation of peak current. Thus, Hm3a increases the time taken for channel desensitization, resulting in a prolonged open state that is observed as an increase in current amplitude consistent with the effect of PcTx1 on ASIC1b (37).

Figure 7.4: Mechanism of action of Hm3a on rASIC1. (A) Representative steady-state desensitization (SSD) current traces in the absence (top) or presence (bottom) of 30 nM Hm3a, currents elicited by pH 6.0 following conditioning (cond.) from pH 7.9 to 7.1. (B) Representative pH- activation current traces in absence (top) or presence (bottom) of 30 nM Hm3a, elicited by pH 5.0 to 7.0. (C) SSD and activation curves of rASIC1a in the absence and presence of 30 nM Hm3a (n = 6).

Statistical comparison of pH50 between control and Hm3a data sets; SSD, p < 0.0001 and F = 379.1;

Activation, p = 0.0219 and F = 5.526. (D) Time course of rASIC1a inhibition by 30 nM Hm3a. The t1/2 was calculated from a single-exponential decay function (n = 6). (E) Recovery of rASIC1a from

104 inhibition by 30 nM Hm3a. The t1/2 was calculated a single-exponential function fit to the data (n = 6). Representative current traces for panels D and E are shown in Figure S7.2. (F) Time constants () of rASIC1b desensitization for control conditions and with increasing concentrations of Hm3a. Dashed line represents a fit of the Hill equation (n = 5). Inset: effect of 1 M Hm3a on rASIC1b with Hm3a potentiated current (dotted line) normalized to control (solid line) peak height to better visualize slowed channel desensitization by Hm3a. Data are mean ± SEM.

7.3.5 Characterizing the molecular interactions involved in subtype and species selectivity of Hm3a on ASIC1

The binding site of PcTx1 on ASIC1a was first studied using a chimeric channel approach (37,53) and has since been extensively defined (27,41,51,52). PcTx1 binds primarily to - helix 5 at the bottom of the acidic pocket and a short stretch of residues following -sheet 3 in the palm of the adjacent subunit (Figure 7.5A and B). The rASIC1a residue Phe350 (Phe352 in hASIC1a) on -helix 5 in the thumb domain is particularly important for interaction with the peptide. Previous studies using diluted Psalmopoeus cambridgei venom (containing PcTx1) or recombinant PcTx1 on the hASIC1a_F352L or rASIC1a_F350A mutants, respectively, showed a total lack of inhibition (51,179). However, shortly after, Sherwood et al., showed evidence that PcTx1 could still bind to the mutant channel (i.e., pre-incubation with 200 nM PcTx1 prevented the potentiating activity of big dynorphin) (179). Given the similarity of Hm3a with PcTx1 in sequence and pharmacology we assume it will share the same binding site.

In order to confirm this, Hm3a was tested on the rASIC1a_F350A mutant. Our studies with higher concentrations (up to 10 M) show that not only can Hm3a still bind, but also that it has an interesting dual functional effect (Figure 7.5C). Initial studies using a conditioning pH of 7.45 (to match that of all other concentration-effect data presented) showed that Hm3a strongly potentiated currents at concentrations over 300 nM. This effect was maximal at 3 M (~2.8-fold increase in current amplitude), and a small decrease in the potentiation efficacy was seen at 10 M. Given that Hm3a inhibits rASIC1a by inducing SSD and that the rASIC1a_F350A pH50 of SSD is shifted by ~0.1 units in the acidic direction compared to wild-type rASIC1a (Figure S7.4), the Hm3a concentration-effect curve was repeated using a conditioning pH of 7.35 (i.e., a conditioning pH that is ~0.2 pH units above the pH50 for SSD, thus equivalent to using pH 7.45 for wild-type rASIC1a). Under these conditions Hm3a, from 100 nM to 1 M, once again potentiated currents (albeit with a much decreased

105 efficacy), however at concentrations greater than 3 M Hm3a induced channel inhibition. This suggests that despite losing the large number of contacts (five crystal contacts) made between Phe350 and PcTx1 (and by similarity, Hm3a) (51) there are still sufficient residual contacts to mediate a functional interaction at higher concentrations. It is thus tempting to speculate that during the potentiating effect (observed with lower peptide concentrations), Hm3a binds with partial occupancy and facilitates channel opening. Whereas, at higher concentrations the increased binding site occupancy (3 per channel) is sufficient to push channels into the desensitised state. Another possible explanation could be that in the presence of the F350A mutation, Hm3a adopts a different orientation when bound, resulting in the complex concentration-dependent functional outcome observed in our study.

ASIC1b is highly expressed in peripheral sensory nerves and appears to play a key role in mediating acid-induced nociception, at least in rodents (18,39). Thus determining the molecular basis for Hm3a potentiation of ASIC1b is an important step to understand potential complications when interpreting in vivo data, and enabling the design of more selective ASIC1 modulators. Thus, we further characterized the potential interactions sites of Hm3a on ASIC1. ASIC1a and ASIC1b are splice isoforms that only differ in the first third of the protein (identical from residue 186 of rASIC1a onwards). The use of a series of chimeras between rASIC1a and rASIC1b revealed a small stretch of residues 167–185 of rASIC1a (19 residues N-terminal to the splice-site) that are sufficient to confer the difference in peptide activity between rASIC1a and rASIC1b (inhibition vs potentiation) (coloured magenta in Figure 7.5B and sequence shown in Figure 7.5D) (37). Interestingly, the only residue difference within the known PcTx1 binding site between human and rat ASIC1a, is Ala178 in the rat channel (valine in hASIC1a) is also part of this small sequence stretch. We examined this region in further detail with rASIC1a point mutants for all PcTx1:cASIC1 crystal contact residues, predicting that several of the non-conserved residues should be responsible for the different subtype selectivity and species-specific pharmacology of Hm3a and PcTx1 at ASIC1 (Figure 7.5D and E).

The IC50 of Hm3a on rASIC1a_A178V (2.1 ± 0.7 nM) was only slightly increased compared to wild-type rASIC1a (Figure 7.5E), which does not account for the 30-fold difference in IC50 on rat and human ASIC1a. We concluded that Ala178 is not strongly involved in channel– peptide interaction, in agreement with molecular dynamics predictions whereby the equivalent residue on cASIC1 (Gln179) does not form persistent interactions with PcTx1 (51). Furthermore, this finding is consistent with a previous study in which the mechanism

106 of action of PcTx1 (i.e., alkaline shift of SSD) was suggested to be responsible for the apparent lower potency of PcTx1 at hASIC1a compared to rodent ASIC1a, as the pH50 SSD of hASIC1a is more acidic than that of the rat or mouse isoform (49).

The IC50 of Hm3a on rASIC1a_H173S (1.8 ± 1.0 nM), rASIC1a_F174Y (1.9 ± 0.9 nM), and rASIC1a_A178P (2.1 ± 1.0 nM) was less than 2-fold different compared to that on wild-type rASIC1a. Consequently, these three residues are not likely to be making important interactions with the peptides for functional activity. In contrast, there was a significant decrease in Hm3a activity on both rASIC1a_R175C and rASIC1a_E177G, with an IC50 of 23.3 ± 1.1 nM and 13.2 ± 1.2 nM, respectively (Figure 7.5E). These mutations were made to make the rASIC1a-binding interface resemble that of rASIC1b. None of these mutations drastically affected the channels sensitivity to protons (in the form of pH-dependence of SSD and activation) (Figure S7.4), thus, any loss in potency is likely due to a loss of important interactions made between the peptide and channel, rather than changes in channel function or peptide mechanism. Consistent with this result, PcTx1:ASIC1 crystal structures and molecular dynamics studies (27,51,52) showed that Arg175 and Glu177 of rASIC1a (Arg176 and Glu178 in cASIC1) make contact with Trp24 and Arg27 of PcTx1, respectively, both major pharmacophore residues (41,51). Although the interaction between PcTx1 and ASIC1a is well characterized, no study has so far demonstrated which individual residues determine subtype specificity between rASIC1a and rASIC1b. Here we show that Hm3a interacts with both ASIC1a and ASIC1b with a ~35-fold difference in potency (but different functional outcomes) and identified Arg175 and Glu177 of rASIC1a as two of the causative residues which can account for the subtype specificity of Hm3a, and most likely PcTx1. This improved understanding of the binding interaction between peptide and channel should aid the design of more specific analogues of these peptides.

107

Figure 7.5: Molecular interactions involved in species and subtype selectivity. (A) Structure of the PcTx1:cASIC1 complex (PDB: 3S3X) viewed from above. The three cASIC1 subunits are colored green, orange, and blue, and the three PcTx1 molecules (grey) are shown bound at the subunit interfaces. (B) Molecular surfaces of the complex indicating the 19-residue region (amino acids 167– 185, highlighted in magenta) that determines functional inhibition of rASIC1a by PcTx1. (C) Concentration-effect curves of Hm3a on rASIC1a_F350A when conditioning the channel at both pH

108 7.45 (black circles) and pH 7.35 (red squares) (n = 6) (illustrated using a lowness fit, see Figure S7.5 for fits with the Hill function for potency determination). (D) Sequence alignment of ASIC subtypes showing the 19 residue functionally important region for PcTx1 (and by similarity Hm3a) (coloured magenta in panel B). Cyan = residues identical in all subtypes, grey = semi-conserved residues (i.e. identical to rASIC1a). Channel residues in rASIC1a that correspond to cASIC1 crystal contacts were mutated in this study, and the residue they were mutated to are bolded. (E) Concentration-effect curves for Hm3a at wild-type rASIC1a and point mutants (n = 6–10). A statistically significant difference from rASIC1a WT in logIC50 inhibition by Hm3a was present for both rASIC1a R175C and E177G mutants, p < 0.0001, represented by *. Data for panel E were performed using a conditioning pH of 7.45. Data are mean ± SEM.

7.3.6 Stability of Hm3a and PcTx1

Given the likely role of ASIC1a in a range of pathological conditions, and the need to use selective pharmacological tools for target validation in vivo, it is important to understand the thermal and biological stability of peptides such as Hm3a and PcTx1. Despite its extensive use as a research tool in vivo, the stability of PcTx1 has never been reported. Therefore, we evaluated the stability of Hm3a and PcTx1 in comparison to the clinically used peptide oxytocin at 55 °C in phosphate-buffered saline at pH 7.4, and in human serum (Figure 7.6). Hm3a proved to have high thermal stability with ~10% loss after 48 h at 55 °C. In comparison, PcTx1 showed slightly more breakdown with ~24% loss after 48 h, and oxytocin showed a near-linear breakdown over time with ~38% of oxytocin remaining at 48 h. Similarly, Hm3a was largely unaffected in human serum with ~87% intact peptide after 48 h. In contrast, degradation of PcTx1 and oxytocin continued over the assay time with ~ 35% and ~40% of peptide respectively remaining after 48 h.

109

Figure 7.6: Comparative stability studies of Hm3a and PcTx1. (A) Thermal stability of Hm3a, PcTx1 and oxytocin at 55 °C (pH 7.4, phosphate-buffered saline). (B) Serum stability of Hm3a, PcTx1 and oxytocin at 37 °C. The percentage of peptide remaining for all conditions was quantified using RP- HPLC (n = 3–5). (C) Example RP-HPLC profile at 214 nm of Hm3a and PcTx1 following incubation in human serum at 0 h (black), 24 h (blue) and 48 h (red). Data are mean ± SEM, and statistical analyses are comparisons between Hm3a and PcTx1 at a given time point (* represents p > 0.05, and ns is not significant).

It is often claimed that ICK peptides are highly stable and resistant to thermal, chemical and enzymatic degradation and this has been demonstrated for several spider venom peptides, ω-Hv1a, ProTx-II, GsMTx-4, and GTx1-15 (180,181). Interestingly, despite their high degree of sequence similarity, Hm3a was more stable than PcTx1 when challenged with either high temperature or human serum. The improved stability of Hm3a over PcTx1 may be due to Hm3a’s fewer charged residues (particularly lysines at the C-terminus), which are common cleavage sites for proteolytic enzymes such as trypsin (182). We conclude that the substantially higher biological stability of Hm3a as compared to PcTx1 makes it a substantially more attractive tool for studying the role of ASICs in biological fluids and in in vivo studies.

110 7.5. CONCLUSIONS

We have identified and characterized a new potent peptide modulator of ASIC1 from the venom of H. maculata, and suspect that tarantula or other spider venoms may harbour even more, potent and selective peptide inhibitors of ASICs. Despite the high sequence and pharmacological similarity of Hm3a and PcTx1, Hm3a appears to be superior in terms of biological stability. Furthermore, through rational engineering of Hm3a, we produced a more potent variant of the peptide making its ASIC1a inhibitory potency on par with PcTx1. We also identified several determinant residues in rASIC1a for ASIC1a/1b selectivity studies. Overall, this study will contribute to the rational engineering and development of ASIC modulators with improved subtype selectivity and facilitate the development of novel ASIC- targeting therapeutic leads and research tools.

111 7.6 SUPPLEMENTARY INFORMATION

Figure S7.1: Example current traces from Xenopus oocyte data corresponding to concentration- response curves presented in Figures 7.1D and 7.3. (A) Native Hm3a isolated from venom inhibits rat ASIC1a (rASIC1a). Recombinant Hm3a activity at (B) rASIC1a, (C) rASIC2a, (D) rASIC3, (E) rASIC1b, (F) rASIC1a/1b heteromers (co-expression of RNAs in oocyte), (G) human ASIC1a (hASIC1a), and (H) hASIC1b. (I) Recombinant Hm3a_P38 mutant activity at rASIC1a. Black boxes indicate peptide application with the corresponding concentration indicated above; grey boxes indicate pH stimulus; and the dashed line indicates current rundown that is independent of peptide application.

112

Figure S7.2: Example current traces from Xenopus oocyte data corresponding to on- and off-rates presented in Figures 7.4D and E. (A and B) Representative traces of pH-evoked rASIC1a current inhibition by 30 nM Hm3a with varying degrees of application time as indicated by labels. (C) Example trace of rASIC1a recovery from inhibition with constant flow of peptide-free solution after 30 nM Hm3a application of 180 s. Currents were evoked every 60 s to assess recovery, and the dashed line indicates current rundown.

Figure S7.3: Potentiation of rASIC1a currents by Hm3a with less than 10 s peptide exposure. (A) pH-stimulated currents are potentiated by application of Hm3a for 5 s, then from 10 s onwards, current inhibition increases with increased peptide exposure time as shown in Figure 4D (n = 4–7). (B) Representative traces of rASIC1a current potentiation by Hm3a application for 5s before pH stimulus with pH 5.0. Dashed line represents control current size and data for panel A are mean ± SEM.

113

Figure S7.4: pH-dependence of (A) steady-state desensitization and (B) activation for wild-type and mutant rASIC1a channels used in this study (n = 6–8). (C) Half- maximal response (pH50) and slope values for steady-state desensitization and activation curves from Hill function fits to the data. Data are mean ± SEM.

Figure S7.5: Equivalent data set from Figure 7.5B showing the concentration-effect curves of Hm3a on the channel mutant rASIC1a_F350A, conditioned at both pH 7.45 (black) and pH 7.35 (red) (n =

6). Unlike Figure 5B, these data have been fit with the Hill function to obtain EC50 values. When calculating the regression for the pH 7.45 data, a range was set between 1 nM and 3 M. Similarly, the data set collected at pH 7.35 has been fit with two equations with different ranges. The potentiation effect has been fit with the data between 1 nM and 1 M, whereas the inhibition with data between 3 M and 10 M. Data are mean ± SEM.

114

CHAPTER 8

Novel conorfamides from Conus austini venom modulate both nicotinic acetylcholine receptors and acid-sensing ion channels

115 Preface to Chapter 8. One of the objectives of this thesis was to discover novel venom peptide modulators of ASICs. This chapter is derived from a submitted manuscript which details the activity of a family of peptides that showed activity at nAChR and ASICs. During my screen of venoms for ASIC activity, no cone snail venom tested yielded significant pharmacology, however the venom of Conus textile was not tested. Irina Vetter, Ai-hua Jin, and Sergio A R González had identified conorfamides from this venom through an initial fluorescence screen of 7 nAChR activity and noted the sequence similarity to the ASIC modulating RFamide peptides. Ai-hua Jin synthetically produced this peptide and I performed the electrophysiology presented here to characterise the nAChR and ASIC pharmacological activity. I contributed to ~65% of the writing, editing, and preparation of figures for the manuscript.

8.1 INTRODUCTION

Cone snail venoms are a rich source of peptides that includes linear forms and disulfide bonded molecules that generally contain 1–5 disulfide bonds (183). The disulfide-poor conopeptides containing none or one disulfide bond have attracted less attention than the cysteine-rich conotoxins (184). Nevertheless, twelve families of these conopeptides have been characterized (184,185), and reported to have broad biological activities. These include inhibition of voltage-gated potassium channels (186), and ligand-gated glutamate receptors (187-191), as well as disruption of cellular membranes (192) and impairment of motor function in mice (193).

Within the conopeptides, the conorfamide family is characterized by an RF-amide motif at the C-terminus of the peptide. The first member (Conorfamide-Sr1) was identified from the venom of Conus spurius in 2001 (194). Subsequently, conorfamide-NPY1 was isolated from Conus betulinus (195), and conorfamide-Vc1 discovered by a proteogenomic annotation strategy from Conus victoriae (196). Intracranial administration of these three conorfamides elicited hyperactivity or paralysis in mice, however, no molecular targets were identified. More recently, Conorfamide-Tx1.1 was shown to elicit pain by enhancing acid-sensing ion channel (ASIC) 3 currents with an EC50 of 3–4 M (95). Interestingly, the conorfamide-Tx1.1 amino acid sequence (RPRFamide) is identical to the last four residues of cono-NPY1. The most recently identified conorfamide-Sr3, also from the venom of Conus spurius, inhibits the

Shaker subtype of voltage-gated potassium (KV) channels with an IC50 of 2.7 M (197). Thus, the subfamily of conorfamides appears to be defined by modest potency at diverse pharmacological targets. 116 ASICs are proton-gated cation channels and members of the degenerin/epithelial sodium channel super family (4). They are widely expressed throughout the body and have been implicated in various physiological and pathological conditions. There are at least six ASIC subtypes (1a, 1b, 2a, 2b, 3, and 4) that form function homo- and hetero-trimeric channels (19). A diverse array of ASIC modulating compounds have been identified, however, the most potent and selective to date have been isolated from animal venoms (33,139).

Nicotinic acetylcholine receptors (nAChRs) are pentameric ligand gated ion channels permeable to Na+ and Ca2+ ions. Homo- or heteromeric channels of  and various accessory subunits combine to produce channels with diverse pharmacological and physiological functions (198). nAChR activity has long been associated with the small (12–19 amino acids), two disulfide bonded -conotoxins (199-201). Indeed -conotoxins have been instrumental in improving our understanding of nAChR pharmacology in vivo and in vitro.

In this work, we report the biochemical and functional characterization of two novel linear conorfamides (conorfamide_As1a and conorfamide_As2a) and their non-amidated counterparts conopeptide_As1b and conopeptide_As2b from the venom of the Mexican cone snail Conus austini. These peptides were identified by high-throughput fluorescence imaging plate reader (FLIPR) assay-guided isolation of α7 nAChR antagonist activity and originally predicted to be new -conotoxins due to their molecular size and nAChR inhibitory activity. Surprisingly, sequence determination revealed activity associated with four linear peptides of the conorfamide family rather than the anticipated -conotoxin family. Pharmacological testing revealed micromolar activity at ASIC1a and ASIC3 and nanomolar activity at both 7 and muscle-type nAChRs for these peptides. This is the first report of conorfamides with dual activity, with the nAChR activity being the most potent molecular target of any conorfamide discovered to date.

8.2 METHODS

8.2.1 Crude venom extraction and fractionation

Adult specimens of C. austini were collected from a depth of ~50-80 m in the Western Gulf of Mexico. They were dissected on ice and the venom was stripped from the duct in 1% formic acid solution, lyophilized and stored at –20º C until use. Crude venom (1 mg) was separated into 72 fractions using a Vydac 218TP C18 column (250 x 4.6 mm, 5 µm) eluted

117 at a flow rate of 0.7 mL/min with 5–90% solvent B over 60 min (solvent A, H2O/0.1% formic acid; solvent B, 90% acetonitrile/0.1% formic acid).

8.2.2 FLIPR-based screen to isolate conopeptides

Activity-guided isolation was carried out using a FLIPR high-throughput Ca2+ assay as previously described in detail (202,203). Human SH-SY5Y cells (European Collection of Cell Cultures) were maintained in RPMI medium (Invitrogen, Australia) supplemented with 15% fetal bovine serum and L-glutamine and passaged every 3–5 days using 0.25% trypsin/EDTA (Invitrogen). Cells were plated at a density of 35,000–50,000 cells per well on 384-well, black-walled imaging plates and cultured for 48 h. Fluorescent responses (excitation 470–495 nm; emission 515–575 nm) were assessed using the FLIPRTetraPlus fluorescent plate reader (Molecular Devices) after 30-min incubation with Calcium 4 No Wash dye (Molecular Devices) diluted in physiological salt solution (PSS; composition in mM: NaCl 140, glucose 11.5, KCl 5.9, MgCl2 1.4, NaH2PO4 1.2, NaHCO3 5, CaCl2 1.8, HEPES 10). The equivalent of 50 µg crude venom, resuspended from lyophilised samples in PSS, was added 5 min prior to stimulation of endogenously expressed α7 nAChR in the presence of the allosteric modulator PNU120596 (10 µM). To assess activity at α7 nAChR, SH-SY5Y cells were stimulated with choline (30 µM) and fluorescence responses monitored for 300 reads (1 read/s). Responses were normalized to baseline using ScreenWorks 3.2.0.14 (Molecular Devices).

8.2.3 Chemistry-based sequencing

Ambiguous amino acid sequences from MSMS-based peptide sequencing were resolved by automated Edman degradation using an Applied Biosystems 494 Procise Protein sequencing System (Australian Proteome Analysis Facility). Conorfamide_As1a was loaded onto a precycled, Biobrene-treated disc and subjected up to 18 cycles of Edman N-terminal sequencing.

8.2.4 Peptide synthesis and co-elution

Peptides were assembled using standard Fmoc chemistry on a Symphony (Protein Technologies Inc.) automated synthesizer on a Fmoc-Rink-amide resin for the amide peptides and a preloaded Fmoc-L-Phe-Wang polystyrene resin for the acid peptides (0.1 mmol scale). Fmoc deprotection was accomplished by treatment with 30% piperidine / DMF (1 x 1.5 min then 1 x 4 min). Fmoc amino acids were coupled with HBTU / DIEA (1:1:1) using

118 a 5-fold excess relative to resin loading. The following side-chain protecting groups were used: Arg(Pbf) and Lys(Boc). Peptides were cleaved from the resin and side-chain protecting groups removed by treatment with 95% TFA / 2.5% TIPS / 2.5% H2O for 2 h at room temperature. Following removal of most of the cleavage solvent under a stream of nitrogen, crude peptides were precipitated with the addition of cold Et2O. The precipitates were washed with Et2O then redissolved in 0.1% TFA / 50% MeCN / H2O, lyophilized and HPLC purified to >95% purity with a final yield of ~6%. Co-elutions were carried out on an ABSCIEX QSTAR Pulsar mass spectrometer. The native F36 and F37 and the synthetic peptides were subjected to LC-ESI-MS at 1% B/min gradient. The LC separation was achieved using a Thermo C18 4.6 x 150 mm column. Extracted ion chromatographs were compared.

8.2.5 Data analyses

Data are presented as mean ± standard error and calculated half-maximal values as 95% CI. Statistical comparisons for the sustained current of ASICs were performed using multiple unpaired t-tests with Holm-Sidak correction for multiple comparisons. P < 0.05 was considered statistically significant, with exact P values shown.

8.3 RESULTS

8.3.1 Peptide isolation and sequencing from C. austini venom

C. austini venom was separated by RP-HPLC into 72 fractions, which were then tested for activity against 7 nAChR endogenously expressed in SH SY5Y cells using the FLIPR platform (Figure 8.1). The inhibitory effect was greatest for fractions 36 and 37 (F36 and F37) which were selected for further characterization.

119

Figure 8.1: Assay guided fractionation of Conus austini venom. RP-HPLC chromatogram of venom with active fractions shaded in grey. Inset above chromatogram shows the degree of inhibition of each fraction tested in an 7 nAChR screen on the FLIPR, active fractions are highlighted in red.

MALDI TOF-MS analysis revealed two masses each in both F36 and F37 with four low molecular weight peptides detected (M+H+ 1631.97, 1659.97, 1632.92 and 1660.92). The masses correspond to peptides 13–15 residues long based on the average mass of an amino acid (110 Da). Due to their small sizes and inhibition of 7 nAChRs, the peptides were predicted to be -conotoxins. Surprisingly, no mass shift occurred upon attempted reduction and alkylation of the peptides, indicating that F36 and F37 contain only linear peptides with no disulfide bonds. The peptides were then subjected to de novo MS-based sequencing. Taking advantage of the small size of the four peptides, sufficient fragments and good coverage were obtained for each of the peptides, allowing almost complete sequence characterization except for the ambiguity of Ile/Leu (Figure 8.2A). Sequences were confirmed by combining the MS data with information from Edman degradation (Figure 8.2B). F36 contained two C-terminally amidated peptides containing an FPRF motif, and thus were named Conorfamide_As1a and Conorfamide_As2a. C-terminal acid versions of the F36 peptides were present in F37 and were called Conopeptide_As1b and Conopeptide_As2b, respectively.

120

Figure 8.2: Peptide sequence determination (A) De novo sequencing by MALDI-MSMS of Conorfamide_As1a and (B) sequences of the four identified peptides.

8.3.2 Peptide synthesis and co-elution

The four peptides were chain assembled using standard Fmoc chemistry. TFA / TIPS / H2O cleavage produced one major product for each of the peptides. A single dominant product of expected mass was obtained for each of the four peptides. Synthetic peptides co-eluted with native material supporting the correct sequence identification (Figure 8.3). NMR and CD analysis showed a lack of well-defined secondary structure (data not shown).

Figure 8.3: Synthesis of peptides and co-elution with native peptides for (A) Conorfamide_As1a and (B) Conopeptide_As1b. 121 8.3.3 Pharmacology at ASICs

RFamide peptides are known to modulate the pH-gated current of ASICs in the micromolar range. Consequently, all four peptides identified here were screened across homomeric rat ASIC1a, 1b, 2a, and 3 channels at 50 M (Figure S8.1). None of the peptides had activity at ASIC1b or 2a. In contrast, As1a and As2a had varied activities at ASIC1a and 3. The amidated peptides As1a and As2a altered the desensitization of ASIC1a (Figure 8.4A), and to a lesser degree ASIC3 currents (Figure 8.4B), resulting in a sustained opening of channels in the presence of low pH. Strikingly, the most significant difference observed was the effect on peak current between the single point mutants As1a and As2a at rASIC1a. As1a, up to 200 M, had little effect on peak currents, whereas As2a potentiated ASIC1a with an EC50 of 10.9 M (Figure 8.4C; 95% confidence interval (CI) pIC50 5.18–4.63 M). Both peptides were equipotent in their inhibition of desensitization (Figure 8.4D). The presence of the C-terminal amidation is crucial for the observed effects on both sustained currents and peak current as the free acid variants were largely inactive, with the exception of some inhibition of ASIC1a by As1b (Figure 8.4A).

Figure 8.4: ASIC activity. Conorfamide and conopeptide (50 M) activities for rat (A) ASIC1a and (B) ASIC3. Example traces are shown, as well as inhibition of peak currents and effect on sustained current. Black line represents control currents and red line represents in the presence of peptides. Concentration-effect curves for As1a and As2a for rat ASIC1a (C) inhibition of peak current and (D) 122 effect on sustained current revealed the single point mutation to be important for the ASIC interaction. Data are mean  SEM and represent n  5. Scale bar is shown where the abscissa scale is 3 s and ordinate scale is 200 nA. P values are shown and “ns” is not significant.

8.3.4 Pharmacology at nAChR

The peptides identified here were isolated based on their activity in an 7 nAChR screen using a fluorescent Ca2+ assay. To determine the potency of the four peptides, concentration-effect curves were established with synthetic material using Xenopus oocytes expressing human 7 nAChR (Figure 8.5A). Both amidated variants (As1a and As2a) were equipotent inhibitors of 7 nAChR with an IC50 of ~4 M (95% CI pIC50 (M): As1a, 5.42– 5.21; As2a, 5.50–5.34). The respective free acid variants were slightly less potent (95% CI pIC50 (M): As1b, 5.19–4.98; As2b, 5.24–5.08). There was a considerable difference in the off-rate when comparing the amide and acid variants (Figure 8.5B). Over 10 minutes of continuous washing, the amide variants recovered only ~10% of current, whereas the free- acid variants had completely washed off within 5 minutes. Conorfamide activity at muscle- type nAChR was also tested in the oocyte assay and showed nanomolar inhibition (Figure

8.5C). The amidated As1a and As2a were equipotent with an IC50 of 212–245 nM, and the free acid variants were again ~3 fold less potent (95% CI pIC50 (M): As1a, 6.75–6.48; As2a, 6.85–6.50; As1b, 6.22–5.92; As2b, 6.30–6.09). Notably, all four peptides washed off the muscle-type channel within 2 minutes (i.e. at the first stimulation after highest concentration tested; data not shown).

Figure 8.5: nAChR activity. (A) Concentration-effect curve (n = 5) and (B) washout at human 7 nAChR expressed in oocytes (n = 3). (C) Concentration-effect curve at human muscle-type nAChR expressed in oocytes (n = 4). As1a, filled black circle; As1b, open black circle; As2a, filled red square; As2b, open red square. All data are mean  SEM. nAChR currents were elicited by a 5 s application of ACh (100 M for α7 and 3 M for muscle-type ()) every 120 s. Peptides were incubated for 50 s for ASICs and 110 s for nAChRs.

123 8.4 DISCUSSION AND CONCLUSIONS

Venom-derived peptides have attracted substantial research interest due to their potential as therapeutics and valuable pharmacological tools (204-206). The treasure trove of bioactive peptides within cone snail venoms have been gradually unravelled over the last 40 years, with current venomic approaches accelerating biodiscovery (207,208) . Unlike many conotoxin families, conorfamides have not been commonly identified, with only seven examples reported prior to this work (Figure 8.6). Likewise, the pharmacology of conorfamides is poorly characterized, with only two molecular targets having been identified to date: the ASICs targeted by conorfamide-Tx peptides, and Shaker potassium channels targeted by conorfamide-Sr3. With the exception of the RFamide motif at the C-terminus, only the presence of an Arg or Phe residue at the fourth last position is well conserved across the known conorfamides. This degree of sequence variability is consistent with the diverse pharmacology of the conorfamides. Our study confirms ASICs as a target of this peptide family and extends the known biological activity to include subtypes of the nicotinic acetylcholine receptor family.

Figure 8.6: Sequence alignment of the conorfamide family. Background colouring indicates residue conservation.

Interestingly, the C-terminal RFamide motif has been identified in at least five neuropeptide families (209). Similar to the conorfamides, these peptides show great N-terminal sequence diversity, which is reflected by the wide range of biological activities described. The first RFamide isolated was the tetrapeptide FMRFa from a clam that was described as cardioexcitatory (210). Since this discovery, FMRFa and related peptides have been shown to directly gate the FMRFa-gated sodium channels (FaNaCs), which are closely related to ASICs (211). Indeed, FMRFa and related peptides have been shown to modulate proton- induced ASIC1 and ASIC3 currents with varying degrees of potency by slowing channel desensitization and inducing a sustained current in the presence of low pH (212).

124 This pharmacological profile was observed at ASIC3 for the conorfamide-Tx peptides, where the shortest variant (RPRFa) was found to be the most potent. RPRFa also potentiated ASIC3 peak currents, but had no effect on ASIC1. RPRFa was recently shown to stabilize the ASIC3 open state and slow the transition to the desensitized state via effects on the ASIC nonproton ligand sensing domain (96). The conorfamides identified here end with the FPRFa amino acid sequence, and agree with the body of work showing that this motif can positively modulate ASIC1a and ASIC3. Recent work showed FRRFa binds to the central vestibule and its side cavities (in the lower palm domain) of ASIC1a to prevent desensitization, a similar binding region and mechanism to RPRFa at ASIC3 (213). The conservative K3R mutation between As1a and As2a that introduced potentiation of ASIC1a here shows that the extended N-terminus in these peptides can be important for molecular target recognition and introducing broader pharmacology. It is thought that binding of the C- terminal RFamide motif to the lower palm/central vestibule alters desensitization kinetics of ASICs. More potent activity of RFamides at ASICs has been reported for shorter compared to longer peptide sequences. We hypothesize that the N-terminus of longer conorfamides may be able to reach more distant regions of the central vestibule and alter the activity profile.

RFamides now join the small molecule GMQ as ASIC modulators to bind to the lower palm domain of ASICs, further validated the importance of this region to ASIC gating. Interestingly, early studies showed a similar region of FaNaCs to be the recognition site for FMRFa (214). A second distinct binding region in the upper finger domain of FaNaCs has been recently reported for FMRFa activity (215,216). This region is known to be important for FaNaC gating, and is also much closer to the acidic pocket of ASICs, the area that initiates the activation cascade for proton-gating. It could be that for ASICs, there are two RFamide binding sites. Shorter sequences may enter the central vestibule through the channel fenestrations and longer sequences bind the finger for activity. The emergence of several ASIC modulating RFamides brings exciting potential to use these peptides to study ASIC gating states that have previously been difficult to stabilize.

Conorfamides have not previously been shown to modulate nAChRs, and the nanomolar potency identified here broadens the scope of channel families targeted by conorfamides. It suggests that these peptides could have evolved to target nAChRs. Indeed, it is conceivable that previous biological activity observed for conorfamide-Vc1 and -Sr1 that lead to hyperactivity or incapacitation in mice, respectively, could in part be due to ASIC, FaNaCs,

125 and/or nAChR modulation. Future work performing a broader screen of various conorfamides with a panel of nAChRs and ASIC/FaNaCs may reveal additional potent interacting partners. Targeting FaNaCs as a defense mechanism and/or for prey capture in mollusk hunting cone snails seems ecologically plausible and an area worthy of further investigation.

In summary, in this work we have isolated two novel conorfamides and their non-amidated counterparts from the venom of C. austini, where the amidated variants target both ASICs and nAChRs. This is the first work showing conorfamides target nAChR, and this molecular target is the most potent for any conorfamide discovered to date. This broadens the scope of ion channels targeted by conorfamides and may reveal an important role of this channel family in conorfamide biology.

8.5 SUPPLEMENTARY INFORMATION

Figure S8.1: Full data set for peptides (50 M) screened across rat (A) ASIC1a, (B) ASIC1b, (C) ASIC2a, and (D) ASIC3. Left graph is inhibition of peak current and right graph effect on sustained current for each ASIC subtype. Error bars represent SEM and individual data points are shown for each n-number (n  5). ASIC currents were elicited by a 5 s drop in pH from 7.45 to 6.3, 5.5, 4.5, and 6.3 (for ASIC1a, ASIC1b, ASIC2a, and ASIC3, respectively) every 60 s.

126

CHAPTER 9

Conclusions and future directions

127 The ultimate aim of this thesis was to better understand the interaction between acid-sensing ion channels and their venom peptide modulators. The work presented herein details the pharmacological characterisation of several known ASIC modulators and the discovery of two novel ASIC ligands. As discussed below, together these have expanded our knowledge of ASIC tools by providing valuable information to allow for their better use in future studies, and will facilitate the design of analogues with improved selectivity for use in target validation studies and may have potential therapeutic value.

Animal venoms have provided researchers with a treasure trove of potent molecules to study ion channels. As reviewed in Chapter 1 ASICs are no exception, with venom peptides proving crucial in deciphering ASIC structure-function and involvement in pathological conditions. The most potent and selective ASIC modulators discovered to date are from animal venoms. However, the number of known venom peptide modulators of ASICs is relatively low when compared to other ion channel families. One would predict that future studies will discover many more potent venom-derived compounds that target ASICs. These compounds may have unique pharmacologies that will deepen our understanding of ASICs physiology, structure and function. It is also important to gain a thorough understanding of the tools we have to maximise their research value. Without a comprehensive understanding of how these peptides interact with ASICs, it can be very easy to misinterpret results from functional and animal studies.

One of the most widely used tools to study ASIC function is the tarantula peptide PcTx1, which was the first potent and selective ASIC1 modulator discovered (34). Nevertheless, the details of how it interacts with several ASIC1 variants to produce different species and subtype-dependent activities had remained far from clear. In Chapter 4 we showed that the different apparent potency of PcTx1 on rat and human ASIC1a is due to variations in their steady-state desensitisation profiles, rather than differences in the PcTx1 binding affinity or mutations in the ASIC1a binding pocket. The different functional outcomes observed depending on PcTx1 application pH is an important point for consideration when using the peptide as a pharmacological tool to study ASICs, particularly in vivo where the exact pH under which the peptide is acting is not known. The application-pH dependent activities we observe may indeed explain why contrasting results have been published from in vivo models of disease using PcTx1 to inhibit ASIC1a function. It would also be interesting to note whether a molecule such as PcTx1 that prevents ASIC1a conductance via stabilising the desensitised state promotes downstream pathways that are different to a molecule that

128 blocks the transition from the closed to open state. Understanding the different mechanism of actions for peptides could help future research that looks to understand the role ASICs play in signalling networks. Alanine-scanning mutagenesis of PcTx1 revealed subtle differences in the pharmacophore at rASIC1a, hASIC1a, and rASIC1b. Future studies that look to further develop PcTx1 analogues to be more selective for a specific ASIC variant and activity will add to the toolkit of pharmacological modulators available to understand ASIC function. Together, our findings have furthered our understanding of the molecular details governing the interaction between PcTx1 and ASIC1 variants.

APETx2 is a 42-residue peptide isolated from the sea anemone Anthopleura elegantissima (62). Despite being discovered ~15 years ago, it remains the most potent ASIC3 inhibitor available, and is a promising therapeutic candidate having shown analgesic activity in several models of pain and most importantly this has been independently confirmed by several groups, including our lab, and a drug company (15,70,72,146). Surprisingly, the APETx2 mechanism of action and binding site have remained elusive, thus Chapter 5 sought to investigate these details. We showed APETx2 binds to the closed state of ASIC3 and prevents the extracellular proton-induced conformational changes that lead to channel opening. Our data suggest binding of APETx2 to the finger domain, a novel binding site for peptide modulators of ASICs. This would be the first example of a larger ASIC ligand binding to the finger domain to exert its effect. Binding here is an fascinating prospect given the growing amount of information revealing the importance of the finger domain in initialising the activation cascade during gating. Indeed FMRFa activation of FaNaCs was recently suggested to be via binding to the finger domain of these closely related channels (215). Nonetheless, the work presented here does not detail the ASIC amino acids involved in this interaction. It will be important in future studies to delineate these details either via extensive mutagenesis studies or potentially a high-resolution complex structure. This information will be required to truly understand the APETx2:ASIC interaction, and to design more selective analogues (as is suggested as possible by the work presented herein) that would provide better tools for studying the role of ASIC3 in physiological and disease states.

Mambalgins are a group of three-finger toxins isolated from black and green mamba snake venoms (18,158). They are the first reported potent inhibitors of rodent ASIC1b and their use, together with siRNA mediated gene silencing, demonstrated that ASIC1b and not ASIC1a is important in peripheral pain sensing in mice (18). In Chapter 6 we show that mambalgins can be successfully produced in E. coli, and have different effects on the gating

129 properties of specific ASIC1 variants. Importantly we show that Ma-3 has only weak inhibitory effects at hASIC1b, and its mechanism of action renders it a potentiator of this subtype under certain conditions. This is extremely important knowledge as mambalgins have been touted as potential pain therapeutics based on their inhibitory activity at ASIC1b and analgesic activity in rodents. We also clarify some of the disagreement in the literature over the binding site interactions of mambalgins with ASIC1a, and show the major binding region to be the middle and lower thumb domain, not the acidic pocket. However, further mutagenesis work will be required on both mambalgins and ASICs to conclude an accurate binding pose.

A more complete picture of the mambalgin:ASIC network of interactions will aid future attempts to design mambalgin analogues that are more potent at human ASIC1b. The availability of a human ASIC1b inhibitor will help determine which ASIC subtype (1a or 1b) is more important for peripheral pain sensing in humans, and whether it is different to rodents. The species dependent activity may also explain the ecological/evolutionary role this peptide plays in the mamba venom. With the initial mambalgin discovery, there was much discussion as to why the mamba snake was producing a peptide that inhibited a channel activated during pain sensation. The very different activity at different species variants highlights that small changes can be important for pharmacology, and that evolutionary/ecological pressures should be studied with known predator-prey ligands and channels.

It is clear that venoms are the most prolific source of potent and selective tools available to study ASICs. Therefore, in an attempt to find new pharmacological tools targeting ASICs, a panel of spider, snake and cone snail venoms were screened for activity at several ASIC subtypes. Chapters 7 and 8 detail the isolation and characterisation of the venom peptides Hm3a from a tarantula, and conorfamide_As1a and 1b from a cone snail. Hm3a is a homologue of PcTx1, with five amino acid substitutions and a truncated C-terminus. The pharmacological profile of Hm3a and PcTx1 is extremely similar, with potent inhibition at ASIC1a and potentiation at ASIC1b. Curiously, Hm3a displayed increased in vivo serum stability over PcTx1, which we propose is due to the fewer charged residues at the C- terminus of Hm3a due to the truncation. This is potentially advantageous for in vivo studies where stability over longer periods is important. This work may lead to the discovery of a family of similar peptides that target ASICs, much like the tens of spider venom peptide modulators of voltage-gated sodium channels. It is of interest to note the company Alomone

130 now offers Hm3a for sale and in their hands it is substantially more potent than the PcTx1 they have for sale. This also demonstrates the broader impact of this discovery.

Cone snail venoms are extremely diverse and a rich source of ion channel modulators, however the first cone snail peptide targeting ASICs was only discovered in 2017 (95). Chapter 8 describes the discovery and characterisation of novel conorfamides, As1a and As1b, from Conus austini venom. Amidated peptide variants altered desensitisation of ASIC1a and 3, and a lysine to arginine mutation introduced ASIC1a peak current potentiation. Given the ASIC activity of this peptide family is relatively new, it is likely RFamides will lead to future work that uses these peptides to better understand gating states, and also be used in animals studies to understand the effect of increasing ASIC activity. Surprisingly, these conorfamides also inhibited 7 and muscle-type nicotinic acetylcholine receptors at nanomolar concentrations. This work broadens the scope of ion channels targeted by conorfamides and may reveal an important role of these channel families in conorfamide biology. One would anticipate cone snails venom to contain several yet to be discovered ASIC modulators, that could prove to have a pharmacological profile yet to be seen. This pharmacology may again be better explained from an evolutionary stand point when studying channels that are relevant to predator-prey relationships.

Overall, the results of this thesis have provided new research tools and represents a significant contribution to our understanding of the pharmacological modulation of ASICs. Furthermore, it lays a solid foundation for the development of more selective tools and potentially novel therapeutic lead molecules for a variety of poorly treated conditions.

131 REFERENCES

1. Garcia-Anoveros, J., Derfler, B., Neville-Golden, J., Hyman, B. T., and Corey, D. P. (1997) BNaC1 and BNaC2 constitute a new family of human neuronal sodium channels related to degenerins and epithelial sodium channels. Proc. Natl. Acad. Sci. USA 94, 1459–1464 2. Price, M. P., Snyder, P. M., and Welsh, M. J. (1996) Cloning and expression of a novel human brain Na+ channel. J. Biol. Chem. 271, 7879-7882 3. Waldmann, R., Champigny, G., Voilley, N., Lauritzen, I., and Lazdunski, M. (1996) The mammalian degenerin MDEG, an amiloride-sensitive cation channel activated by mutations causing neurodegeneration in Caenorhabditis elegans. J. Biol. Chem. 271, 10433-10436 4. Waldmann, R., Champigny, G., Bassilana, F., Heurteaux, C., and Lazdunski, M. (1997) A proton-gated cation channel involved in acid-sensing. Nature 386, 173–177 5. Krishtal, O., and Pidoplichko, V. (1980) A receptor for protons in the nerve cell membrane. Neuroscience 5, 2325-2327 6. Krishtal, O. A., and Pidoplichko, V. I. (1981) A receptor for protons in the membrane of sensory neurons may participate in nociception. Neuroscience 6, 2599-2601 7. Coryell, M. W., Wunsch, A. M., Haenfler, J. M., Allen, J. E., Schnizler, M., Ziemann, A. E., Cook, M. N., Dunning, J. P., Price, M. P., Rainier, J. D., Liu, Z., Light, A. R., Langbehn, D. R., and Wemmie, J. A. (2009) Acid-sensing ion channel-1a in the amygdala, a novel therapeutic target in depression-related behavior. J Neurosci 29, 5381-5388 8. Deval, E., Gasull, X., Noël, J., Salinas, M., Baron, A., Diochot, S., and Lingueglia, E. (2010) Acid-sensing ion channels (ASICs): pharmacology and implication in pain. Pharmacol. Ther. 128, 549–558 9. Wemmie, J. A., Taugher, R. J., and Kreple, C. J. (2013) Acid-sensing ion channels in pain and disease. Nat. Rev. Neurosci. 14, 461–471 10. Bässler, E. L., Ngo-Anh, T. J., Geisler, H. S., Ruppersberg, J. P., and Gründer, S. (2001) Molecular and functional characterization of acid-sensing ion channel (ASIC) 1b. J. Biol. Chem. 276, 33782-33787 11. Lingueglia, E., de Weille, J. R., Bassilana, F., Heurteaux, C., Sakai, H., Waldmann, R., and Lazdunski, M. (1997) A modulatory subunit of acid sensing ion channels in brain and dorsal root ganglion cells. J Biol Chem 272, 29778-29783

132 12. Delaunay, A., Gasull, X., Salinas, M., Noel, J., Friend, V., Lingueglia, E., and Deval, E. (2012) Human ASIC3 channel dynamically adapts its activity to sense the extracellular pH in both acidic and alkaline directions. Proc. Natl. Acad. Sci. USA 109, 13124-13129 13. Lin, S. H., Sun, W. H., and Chen, C. C. (2015) Genetic exploration of the role of acid- sensing ion channels. Neuropharmacology 94, 99-118 14. Coryell, M. W., Ziemann, A. E., Westmoreland, P. J., Haenfler, J. M., Kurjakovic, Z., Zha, X. M., Price, M., Schnizler, M. K., and Wemmie, J. A. (2007) Targeting ASIC1a reduces innate fear and alters neuronal activity in the fear circuit. Biol Psychiatry 62, 1140-1148 15. Izumi, M., Ikeuchi, M., Ji, Q., and Tani, T. (2012) Local ASIC3 modulates pain and disease progression in a rat model of osteoarthritis. J. Biomed. Sci. 19, 77 16. Waldmann, R., Bassilana, F., de Weille, J., Champigny, G., Heurteaux, C., and Lazdunski, M. (1997) Molecular cloning of a non-inactivating proton-gated Na+ channel specific for sensory neurons. J Biol Chem 272, 20975-20978 17. Babinski, K., Le, K. T., and Séguéla, P. (1999) Molecular cloning and regional distribution of a human proton receptor subunit with biphasic functional properties. J. Neurochem. 72, 51-57 18. Diochot, S., Baron, A., Salinas, M., Douguet, D., Scarzello, S., Dabert-Gay, A. S., Debayle, D., Friend, V., Alloui, A., Lazdunski, M., and Lingueglia, E. (2012) Black mamba venom peptides target acid-sensing ion channels to abolish pain. Nature 490, 552-555 19. Jasti, J., Furukawa, H., Gonzales, E. B., and Gouaux, E. (2007) Structure of acid- sensing ion channel 1 at 1.9 A resolution and low pH. Nature 449, 316-323 20. Cristofori-Armstrong, B., Soh, M. S., Talwar, S., Brown, D. L., Griffin, J. D., Dekan, Z., Stow, J. L., King, G. F., Lynch, J. W., and Rash, L. D. (2015) Xenopus borealis as an alternative source of oocytes for biophysical and pharmacological studies of neuronal ion channels. Sci Rep 5, 14763 21. Grunder, S., and Pusch, M. (2015) Biophysical properties of acid-sensing ion channels (ASICs). Neuropharmacology 94, 9-18 22. Hesselager, M., Timmermann, D. B., and Ahring, P. K. (2004) pH Dependency and desensitization kinetics of heterologously expressed combinations of acid-sensing ion channel subunits. J Biol Chem 279, 11006-11015

133 23. Wu, L. J., Duan, B., Mei, Y. D., Gao, J., Chen, J. G., Zhuo, M., Xu, L., Wu, M., and Xu, T. L. (2004) Characterization of acid-sensing ion channels in dorsal horn neurons of rat spinal cord. J. Biol. Chem. 279, 43716-43724 24. Babini, E., Paukert, M., Geisler, H. S., and Grunder, S. (2002) Alternative splicing and interaction with di- and polyvalent cations control the dynamic range of acid- sensing ion channel 1 (ASIC1). J Biol Chem 277, 41597-41603 25. Gonzales, E. B., Kawate, T., and Gouaux, E. (2009) Pore architecture and ion sites in acid-sensing ion channels and P2X receptors. Nature 460, 599-604 26. Baconguis, I., Bohlen, C. J., Goehring, A., Julius, D., and Gouaux, E. (2014) X-ray structure of acid-sensing ion channel 1-snake toxin complex reveals open state of a Na(+)-selective channel. Cell 156, 717-729 27. Baconguis, I., and Gouaux, E. (2012) Structural plasticity and dynamic selectivity of acid-sensing ion channel-spider toxin complexes. Nature 489, 400–405 28. Yoder, N., Yoshioka, C., and Gouaux, E. (2018) Gating mechanisms of acid-sensing ion channels. Nature 555, 397-401 29. Ramaswamy, S. S., MacLean, D. M., Gorfe, A. A., and Jayaraman, V. (2013) Proton- mediated conformational changes in an acid-sensing ion channel. J. Biol. Chem. 288, 35896–35903 30. Vullo, S., Bonifacio, G., Roy, S., Johner, N., Berneche, S., and Kellenberger, S. (2017) Conformational dynamics and role of the acidic pocket in ASIC pH-dependent gating. Proc Natl Acad Sci U S A 114, 3768-3773 31. Baron, A., Diochot, S., Salinas, M., Deval, E., Noel, J., and Lingueglia, E. (2013) Venom toxins in the exploration of molecular, physiological and pathophysiological functions of acid-sensing ion channels. Toxicon 75, 187-204 32. Baron, A., and Lingueglia, E. (2015) Pharmacology of acid-sensing ion channels - Physiological and therapeutical perspectives. Neuropharmacology 94, 19-35 33. Rash, L. D. (2017) Acid-Sensing Ion Channel Pharmacology, Past, Present, and Future. Adv Pharmacol 79, 35-66 34. Escoubas, P., De Weille, J. R., Lecoq, A., Diochot, S., Waldmann, R., Champigny, G., Moinier, D., Menez, A., and Lazdunski, M. (2000) Isolation of a tarantula toxin specific for a class of proton-gated Na+ channels. J Biol Chem 275, 25116-25121 35. Joeres, N., Augustinowski, K., Neuhof, A., Assmann, M., and Grunder, S. (2016) Functional and pharmacological characterization of two different ASIC1a/2a heteromers reveals their sensitivity to the spider toxin PcTx1. Sci Rep 6, 27647

134 36. Sherwood, T. W., Lee, K. G., Gormley, M. G., and Askwith, C. C. (2011) Heteromeric acid-sensing ion channels (ASICs) composed of ASIC2b and ASIC1a display novel channel properties and contribute to acidosis-induced neuronal death. J Neurosci 31, 9723-9734 37. Chen, X., Kalbacher, H., and Grunder, S. (2006) Interaction of acid-sensing ion channel (ASIC) 1 with the tarantula toxin psalmotoxin 1 is state dependent. J Gen Physiol 127, 267-276 38. Smith, R. N., and Gonzales, E. B. (2014) Protons and Psalmotoxin-1 reveal nonproton ligand stimulatory sites in chicken acid-sensing ion channel: Implication for simultaneous modulation in ASICs. Channels 8, 49-61 39. Hoagland, E. N., Sherwood, T. W., Lee, K. G., Walker, C. J., and Askwith, C. C. (2010) Identification of a Calcium Permeable Human Acid-sensing Ion Channel 1 Transcript Variant. J Biol Chem 285, 41852-41862 40. Escoubas, P., Bernard, C., Lambeau, G., Lazdunski, M., and Darbon, H. (2003) Recombinant production and solution structure of PcTx1, the specific peptide inhibitor of ASIC1a proton-gated cation channels. Protein science : a publication of the Protein Society 12, 1332–1343 41. Saez, N. J., Mobli, M., Bieri, M., Chassagnon, I. R., Malde, A. K., Gamsjaeger, R., Mark, A. E., Gooley, P. R., Rash, L. D., and King, G. F. (2011) A dynamic pharmacophore drives the interaction between Psalmotoxin-1 and the putative drug target acid-sensing ion channel 1a. Mol Pharmacol 80, 796-808 42. Jensen, J. E., Cristofori-Armstrong, B., Anangi, R., Rosengren, K. J., Lau, C. H., Mobli, M., Brust, A., Alewood, P. F., King, G. F., and Rash, L. D. (2014) Understanding the molecular basis of toxin promiscuity: the analgesic sea anemone peptide APETx2 interacts with acid-sensing ion channel 3 and hERG channels via overlapping pharmacophores. J Med Chem 57, 9195-9203 43. Chassagnon, I. R., McCarthy, C. A., Chin, Y. K., Pineda, S. S., Keramidas, A., Mobli, M., Pham, V., De Silva, T. M., Lynch, J. W., Widdop, R. E., Rash, L. D., and King, G. F. (2017) Potent neuroprotection after stroke afforded by a double-knot spider-venom peptide that inhibits acid-sensing ion channel 1a. Proc Natl Acad Sci U S A 114, 3750- 3755 44. Mourier, G., Salinas, M., Kessler, P., Stura, E. A., Leblanc, M., Tepshi, L., Besson, T., Diochot, S., Baron, A., Douguet, D., Lingueglia, E., and Servent, D. (2016) Mambalgin-1 Pain-relieving Peptide, Stepwise Solid-phase Synthesis, Crystal

135 Structure, and Functional Domain for Acid-sensing Ion Channel 1a Inhibition. J Biol Chem 291, 2616-2629 45. Schroeder, C. I., Rash, L. D., Vila-Farres, X., Rosengren, K. J., Mobli, M., King, G. F., Alewood, P. F., Craik, D. J., and Durek, T. (2014) Chemical Synthesis, 3D Structure, and ASIC Binding Site of the Toxin Mambalgin-2. Angew Chem Int Edit 53, 1017-1020 46. Osmakov, D. I., Kozlov, S. A., Andreev, Y. A., Koshelev, S. G., Sanamyan, N. P., Sanamyan, K. E., Dyachenko, I. A., Bondarenko, D. A., Murashev, A. N., Mineev, K. S., Arseniev, A. S., and Grishin, E. V. (2013) Sea anemone peptide with uncommon beta-hairpin structure inhibits acid-sensing ion channel 3 (ASIC3) and reveals analgesic activity. J. Biol. Chem. 288, 23116-23127 47. Skarżyński, T. (1992) Crystal structure of α-dendrotoxin from the green mamba venom and its comparison with the structure of bovine pancreatic trypsin inhibitor. J. Mol. Biol. 224, 671-683 48. Chen, X., Kalbacher, H., and Grunder, S. (2005) The tarantula toxin psalmotoxin 1 inhibits acid-sensing ion channel (ASIC) 1a by increasing its apparent H+ affinity. J Gen Physiol 126, 71-79 49. Sherwood, T. W., and Askwith, C. C. (2008) Endogenous arginine-phenylalanine- amide-related peptides alter steady-state desensitization of ASIC1a. J Biol Chem 283, 1818-1830 50. Salinas, M., Besson, T., Delettre, Q., Diochot, S., Boulakirba, S., Douguet, D., and Lingueglia, E. (2014) Binding Site and Inhibitory Mechanism of the Mambalgin-2 Pain-relieving Peptide on Acid-sensing Ion Channel 1a. J Biol Chem 289, 13363- 13373 51. Saez, N. J., Deplazes, E., Cristofori-Armstrong, B., Chassagnon, I. R., Lin, X., Mobli, M., Mark, A. E., Rash, L. D., and King, G. F. (2015) Molecular dynamics and functional studies define a hot spot of crystal contacts essential for PcTx1 inhibition of acid-sensing ion channel 1a. Br J Pharmacol 172, 4985-4995 52. Dawson, R. J., Benz, J., Stohler, P., Tetaz, T., Joseph, C., Huber, S., Schmid, G., Hugin, D., Pflimlin, P., Trube, G., Rudolph, M. G., Hennig, M., and Ruf, A. (2012) Structure of the acid-sensing ion channel 1 in complex with the gating modifier Psalmotoxin 1. Nat Commun 3, 936

136 53. Salinas, M., Rash, L. D., Baron, A., Lambeau, G., Escoubas, P., and Lazdunski, M. (2006) The receptor site of the spider toxin PcTx1 on the proton-gated cation channel ASIC1a. J Physiol 570, 339-354 54. Li, T., Yang, Y., and Canessa, C. M. (2011) Outlines of the pore in open and closed conformations describe the gating mechanism of ASIC1. Nat. Commun. 2, 399 55. Wemmie, J. A., Price, M. P., and Welsh, M. J. (2006) Acid-sensing ion channels: advances, questions and therapeutic opportunities. Trends Neurosci. 29, 578-586 56. Xiong, Z. G., Zhu, X. M., Chu, X. P., Minami, M., Hey, J., Wei, W. L., MacDonald, J. F., Wemmie, J. A., Price, M. P., Welsh, M. J., and Simon, R. P. (2004) Neuroprotection in ischemia: blocking calcium-permeable acid-sensing ion channels. Cell 118, 687-698 57. McCarthy, C. A., Rash, L. D., Chassagnon, I. R., King, G. F., and Widdop, R. E. (2015) PcTx1 affords neuroprotection in a conscious model of stroke in hypertensive rats via selective inhibition of ASIC1a. Neuropharmacology 99, 650-657 58. Pignataro, G., Simon, R. P., and Xiong, Z.-G. (2007) Prolonged activation of ASIC1a and the time window for neuroprotection in cerebral ischaemia. Brain 130, 151-158 59. Duan, B., Wang, Y. Z., Yang, T., Chu, X. P., Yu, Y., Huang, Y., Cao, H., Hansen, J., Simon, R. P., Zhu, M. X., Xiong, Z. G., and Xu, T. L. (2011) Extracellular spermine exacerbates ischemic neuronal injury through sensitization of ASIC1a channels to extracellular acidosis. J. Neurosci. 31, 2101–2112 60. Er, S. Y., Cristofori-Armstrong, B., Escoubas, P., and Rash, L. D. (2017) Discovery and molecular interaction studies of a highly stable, tarantula peptide modulator of acid-sensing ion channel 1. Neuropharmacology 127, 185-195 61. Bohlen, C. J., Priel, A., Zhou, S., King, D., Siemens, J., and Julius, D. (2010) A bivalent tarantula toxin activates the capsaicin receptor, TRPV1, by targeting the outer pore domain. Cell 141, 834-845 62. Diochot, S., Baron, A., Rash, L. D., Deval, E., Escoubas, P., Scarzello, S., Salinas, M., and Lazdunski, M. (2004) A new sea anemone peptide, APETx2, inhibits ASIC3, a major acid-sensitive channel in sensory neurons. EMBO J 23, 1516-1525 63. Deval, E., Noël, J., Gasull, X., Delaunay, A., Alloui, A., Friend, V., Eschalier, A., Lazdunski, M., and Lingueglia, E. (2011) Acid-sensing ion channels in postoperative pain. J. Neurosci. 31, 6059-6066

137 64. Chagot, B., Escoubas, P., Diochot, S., Bernard, C., Lazdunski, M., and Darbon, H. (2005) Solution structure of APETx2, a specific peptide inhibitor of ASIC3 proton- gated channels. Protein Sci 14, 2003-2010 65. Peigneur, S., Beress, L., Moller, C., Mari, F., Forssmann, W. G., and Tytgat, J. (2012) A natural point mutation changes both target selectivity and mechanism of action of sea anemone toxins. Faseb J 26, 5141-5151 66. Blanchard, M. G., Rash, L. D., and Kellenberger, S. (2012) Inhibition of voltage-gated Na(+) currents in sensory neurones by the sea anemone toxin APETx2. Br. J. Pharmacol. 165, 2167-2177 67. Anangi, R., Rash, L. D., Mobli, M., and King, G. F. (2012) Functional Expression in Escherichia coli of the Disulfide-Rich Sea Anemone Peptide APETx2, a Potent Blocker of Acid-Sensing Ion Channel 3. Mar Drugs 10, 1605-1618 68. Rahman, T., and Smith, E. S. (2014) In silico assessment of interaction of sea anemone toxin APETx2 and acid sensing ion channel 3. Biochem. Biophys. Res. Commun. 450, 384-389 69. Kozlov, S. A., Osmakov, D. I., Andreev Ia, A., Koshelev, S. G., Gladkikh, I. N., Monastyrnaia, M. M., Kozlovskaia, E. P., and Grishin, E. V. (2012) [Polypeptide toxin from sea anemone inhibiting proton-sensitive channel ASIC3]. Bioorg. Khim. 38, 653- 659 70. Deval, E., Noel, J., Lay, N., Alloui, A., Diochot, S., Friend, V., Jodar, M., Lazdunski, M., and Lingueglia, E. (2008) ASIC3, a sensor of acidic and primary inflammatory pain. EMBO J 27, 3047-3055 71. Ikeuchi, M., Kolker, S. J., and Sluka, K. A. (2009) Acid-sensing ion channel 3 expression in mouse knee joint afferents and effects of carrageenan-induced arthritis. J. Pain 10, 336-342 72. Karczewski, J., Spencer, R. H., Garsky, V. M., Liang, A., Leitl, M. D., Cato, M. J., Cook, S. P., Kane, S., and Urban, M. O. (2010) Reversal of acid-induced and inflammatory pain by the selective ASIC3 inhibitor, APETx2. Brit J Pharmacol 161, 950-960 73. Ikeuchi, M., Kolker, S. J., Burnes, L. A., Walder, R. Y., and Sluka, K. A. (2008) Role of ASIC3 in the primary and secondary hyperalgesia produced by joint inflammation in mice. Pain 137, 662-669

138 74. Yagi, J., Wenk, H. N., Naves, L. A., and McCleskey, E. W. (2006) Sustained currents through ASIC3 ion channels at the modest pH changes that occur during myocardial ischemia. Circulation research 99, 501-509 75. Jensen, J. E., Mobli, M., Brust, A., Alewood, P. F., King, G. F., and Rash, L. D. (2012) Cyclisation Increases the Stability of the Sea Anemone Peptide APETx2 but Decreases Its Activity at Acid-Sensing Ion Channel 3. Mar Drugs 10, 1511-1527 76. Bohlen, C. J., Chesler, A. T., Sharif-Naeini, R., Medzihradszky, K. F., Zhou, S., King, D., Sanchez, E. E., Burlingame, A. L., Basbaum, A. I., and Julius, D. (2011) A heteromeric Texas coral snake toxin targets acid-sensing ion channels to produce pain. Nature 479, 410-U167 77. Morgan, D. L., Borys, D. J., Stanford, R., Kjar, D., and Tobleman, W. (2007) Texas coral snake (Micrurus tener) bites. South Med. J. 100, 152-156 78. Nishioka, S. A., Silveira, P. V., and Menzes, L. B. (1993) Coral snake bite and severe local pain. Ann. Trop. Med. Parasitol. 87, 429-431 79. Kellenberger, S., Gautschi, I., and Schild, L. (1999) A single point mutation in the pore region of the epithelial Na+ channel changes ion selectivity by modifying molecular sieving. Proc. Natl. Acad. Sci. USA 96, 4170-4175 80. Carattino, M. D., and Della Vecchia, M. C. (2012) Contribution of residues in second transmembrane domain of ASIC1a protein to ion selectivity. J. Biol. Chem. 287, 12927-12934 81. Kellenberger, S., Auberson, M., Gautschi, I., Schneeberger, E., and Schild, L. (2001) Permeability properties of ENaC selectivity filter mutants. J. Gen. Physiol. 118, 679- 692 82. Yang, L., and Palmer, L. G. (2014) Ion conduction and selectivity in acid-sensing ion channel 1. J. Gen. Physiol. 144, 245-255 83. Li, T., Yang, Y., and Canessa, C. M. (2010) Leu85 in the beta1-beta2 linker of ASIC1 slows activation and decreases the apparent proton affinity by stabilizing a closed conformation. J. Biol. Chem. 285, 22706-22712 84. Roy, S., Boiteux, C., Alijevic, O., Liang, C., Bernèche, S., and Kellenberger, S. (2013) Molecular determinants of desensitization in an ENaC/degenerin channel. FASEB J. 27, 5034-5045 85. Springauf, A., Bresenitz, P., and Gründer, S. (2011) The interaction between two extracellular linker regions controls sustained opening of acid-sensing ion channel 1. J. Biol. Chem. 286, 24374-24384

139 86. Alijevic, O., and Kellenberger, S. (2012) Subtype-specific modulation of acid-sensing ion channel (ASIC) function by 2-guanidine-4-methylquinazoline. J. Biol. Chem. 287, 36059-36070 87. Young, G. T., Gutteridge, A., Fox, H. D., Wilbrey, A. L., Cao, L., Cho, L. T., Brown, A. R., Benn, C. L., Kammonen, L. R., Friedman, J. H., Bictash, M., Whiting, P., Bilsland, J. G., and Stevens, E. B. (2014) Characterizing human stem cell-derived sensory neurons at the single-cell level reveals their ion channel expression and utility in pain research. Mol. Ther. 22, 1530-1543 88. Diochot, S., Alloui, A., Rodrigues, P., Dauvois, M., Friend, V., Aissouni, Y., Eschalier, A., Lingueglia, E., and Baron, A. (2016) Analgesic effects of mambalgin peptide inhibitors of acid-sensing ion channels in inflammatory and neuropathic pain. Pain 157, 552-559 89. Sun, D. M., Yu, Y., Xue, X. B., Pan, M., Wen, M., Li, S. Y., Qu, Q., Li, X. R., Zhang, L. H., Li, X. M., Liu, L., Yang, M. J., and Tian, C. L. (2018) Cryo-EM structure of the ASIC1 a-mambalgin-1 complex reveals that the peptide toxin mambalgin-1 inhibits acid-sensing ion channels through an unusual allosteric effect. Cell Discov 4 90. Pan, M., He, Y., Wen, M., Wu, F., Sun, D., Li, S., Zhang, L., Li, Y., and Tian, C. (2014) One-pot hydrazide-based native chemical ligation for efficient chemical synthesis and structure determination of toxin Mambalgin-1. Chem. Commun. ( 50, 5837-5839 91. Mazzuca, M., Heurteaux, C., Alloui, A., Diochot, S., Baron, A., Voilley, N., Blondeau, N., Escoubas, P., Gelot, A., Cupo, A., Zimmer, A., Zimmer, A. M., Eschalier, A., and Lazdunski, M. (2007) A tarantula peptide against pain via ASIC1a channels and opioid mechanisms. Nat. Neurosci. 10, 943-945 92. Osmakov, D. I., Koshelev, S. G., Andreev, Y. A., Dyachenko, I. A., Bondarenko, D. A., Murashev, A. N., Grishin, E. V., and Kozlov, S. A. (2016) Conversed mutagenesis of an inactive peptide to ASIC3 inhibitor for active sites determination. Toxicon 116, 11-16 93. Rodríguez, A. A., Salceda, E., Garateix, A. G., Zaharenko, A. J., Peigneur, S., López, O., Pons, T., Richardson, M., Díaz, M., Hernández, Y., Ständker, L., Tytgat, J., and Soto, E. (2014) A novel sea anemone peptide that inhibits acid-sensing ion channels. Peptides 53, 3-12 94. Báez, A., Salceda, E., Fló, M., Graña, M., Fernández, C., Vega, R., and Soto, E. (2015) α-Dendrotoxin inhibits the ASIC current in dorsal root ganglion neurons from rat. Neurosci. Lett. 606, 42-47

140 95. Reimers, C., Lee, C. H., Kalbacher, H., Tian, Y., Hung, C. H., Schmidt, A., Prokop, L., Kauferstein, S., Mebs, D., Chen, C. C., and Grunder, S. (2017) Identification of a cono-RFamide from the venom of Conus textile that targets ASIC3 and enhances muscle pain. Proc Natl Acad Sci U S A 114, E3507-E3515 96. Reiners, M., Margreiter, M. A., Oslender-Bujotzek, A., Rossetti, G., Grunder, S., and Schmidt, A. (2018) The Conorfamide RPRFa Stabilizes the Open Conformation of Acid-Sensing Ion Channel 3 via the Nonproton Ligand-Sensing Domain. Mol Pharmacol 94, 1114-1124 97. Escoubas, P., Sollod, B., and King, G. F. (2006) Venom landscapes: mining the complexity of spider venoms via a combined cDNA and mass spectrometric approach. Toxicon 47, 650-663 98. Dutertre, S., Jin, A. H., Kaas, Q., Jones, A., Alewood, P. F., and Lewis, R. J. (2013) Deep venomics reveals the mechanism for expanded peptide diversity in cone snail venom. Mol Cell Proteomics 12, 312-329 99. Baneyx, F., and Mujacic, M. (2004) Recombinant protein folding and misfolding in Escherichia coli. Nat Biotechnol 22, 1399-1408 100. Choi, J. H., and Lee, S. Y. (2004) Secretory and extracellular production of recombinant proteins using Escherichia coli. Appl Microbiol Biotechnol 64, 625-635 101. Salinas, G., Pellizza, L., Margenat, M., Flo, M., and Fernandez, C. (2011) Tuned Escherichia coli as a host for the expression of disulfide-rich proteins. Biotechnol J 6, 686-699 102. Klint, J. K., Senff, S., Saez, N. J., Seshadri, R., Lau, H. Y., Bende, N. S., Undheim, E. A., Rash, L. D., Mobli, M., and King, G. F. (2013) Production of recombinant disulfide-rich venom peptides for structural and functional analysis via expression in the periplasm of E. coli. PLoS One 8, e63865 103. Kapust, R. B., Tozser, J., Copeland, T. D., and Waugh, D. S. (2002) The P1' specificity of tobacco etch virus protease. Biochem Biophys Res Commun 294, 949- 955 104. Saez, N. J., Cristofori-Armstrong, B., Anangi, R., and King, G. F. (2017) A Strategy for Production of Correctly Folded Disulfide-Rich Peptides in the Periplasm of E. coli. Methods Mol Biol 1586, 155-180 105. Fukuyama, Y., Iwamoto, S., and Tanaka, K. (2006) Rapid sequencing and disulfide mapping of peptides containing disulfide bonds by using 1,5-diaminonaphthalene as a reductive matrix. J Mass Spectrom 41, 191-201

141 106. Qi, D., and Scholthof, K. B. (2008) A one-step PCR-based method for rapid and efficient site-directed fragment deletion, insertion, and substitution mutagenesis. J. Virol. Methods 149, 85–90 107. Grandori, R., Struck, K., Giovanielli, K., and Carey, J. (1997) A three-step PCR protocol for construction of chimeric proteins. Protein Eng 10, 1099-1100 108. Gurdon, J. B., Lane, C. D., Woodland, H. R., and Marbaix, G. (1971) Use of frog eggs and oocytes for the study of messenger RNA and its translation in living cells. Nature 233, 177-182 109. Barnard, E. A., Miledi, R., and Sumikawa, K. (1982) Translation of exogenous messenger RNA coding for nicotinic acetylcholine receptors produces functional receptors in Xenopus oocytes. Proceedings of the Royal Society of London. Series B, Containing papers of a Biological character. Royal Society 215, 241-246 110. Goldin, A. L. (2006) Expression of Ion Channels in Xenopus Oocytes. Expression and Analysis of Recombinant Ion Channels: From Structural Studies to Pharmacological Screening 111. Sigel, E. (1990) Use of Xenopus oocytes for the functional expression of plasma membrane proteins. The Journal of membrane biology 117, 201-221 112. Lee, H., Lin, M. C., Kornblum, H. I., Papazian, D. M., and Nelson, S. F. (2014) Exome sequencing identifies de novo gain of function missense mutation in KCND2 in identical twins with autism and seizures that slows potassium channel inactivation. Human molecular genetics 23, 3481-3489 113. Simons, C., Rash, L. D., Crawford, J., Ma, L., Cristofori-Armstrong, B., Miller, D., Ru, K., Baillie, G. J., Alanay, Y., Jacquinet, A., Debray, F. G., Verloes, A., Shen, J., Yesil, G., Guler, S., Yuksel, A., Cleary, J. G., Grimmond, S. M., McGaughran, J., King, G. F., Gabbett, M. T., and Taft, R. J. (2014) Mutations in the voltage-gated potassium channel gene KCNH1 cause Temple-Baraitser syndrome and epilepsy. Nat Genet 114. Abu-Daya, A., Khokha, M. K., and Zimmerman, L. B. (2012) The hitchhiker's guide to Xenopus genetics. Genesis 50, 164-175 115. Bertin, A., Hanna, P., Otarola, G., Fritz, A., Henriquez, J. P., and Marcellini, S. (2014) Cellular and molecular characterization of a novel primary osteoblast culture from the vertebrate model organism Xenopus tropicalis. Histochemistry and cell biology 116. de Robertis, E. M., and Black, P. (1979) Hybrids of Xenopus laevis and Xenopus borealis express proteins from both parents. Developmental biology 68, 334-339

142 117. Goda, T., Abu-Daya, A., Carruthers, S., Clark, M. D., Stemple, D. L., and Zimmerman, L. B. (2006) Genetic screens for mutations affecting development of Xenopus tropicalis. PLoS genetics 2, e91 118. Li, B., Russell, S. C., Zhang, J., Hedrick, J. L., and Lebrilla, C. B. (2011) Structure determination by MALDI-IRMPD mass spectrometry and exoglycosidase digestions of O-linked oligosaccharides from Xenopus borealis egg jelly. Glycobiology 21, 877- 894 119. Lloyd, R. E., Foster, P. G., Guille, M., and Littlewood, D. T. (2012) Next generation sequencing and comparative analyses of Xenopus mitogenomes. BMC genomics 13, 496 120. Evans, B. J., Kelley, D. B., Tinsley, R. C., Melnick, D. J., and Cannatella, D. C. (2004) A mitochondrial DNA phylogeny of African clawed frogs: phylogeography and implications for polyploid evolution. Mol Phylogenet Evol 33, 197-213 121. Gui, J., Liu, B., Cao, G., Lipchik, A. M., Perez, M., Dekan, Z., Mobli, M., Daly, N. L., Alewood, P. F., Parker, L. L., King, G. F., Zhou, Y., Jordt, S. E., and Nitabach, M. N. (2014) A tarantula-venom peptide antagonizes the TRPA1 nociceptor ion channel by binding to the S1-S4 gating domain. Current biology : CB 24, 473-483 122. Weber, W. M., Liebold, K. M., Reifarth, F. W., and Clauss, W. (1995) The Ca(2+)- induced leak current in Xenopus oocytes is indeed mediated through a Cl- channel. The Journal of membrane biology 148, 263-275 123. Weber, W. M., Liebold, K. M., Reifarth, F. W., Uhr, U., and Clauss, W. (1995) Influence of extracellular Ca2+ on endogenous Cl- channels in Xenopus oocytes. Pflugers Archiv : European journal of physiology 429, 820-824 124. Middleton, R. E., Warren, V. A., Kraus, R. L., Hwang, J. C., Liu, C. J., Dai, G., Brochu, R. M., Kohler, M. G., Gao, Y. D., Garsky, V. M., Bogusky, M. J., Mehl, J. T., Cohen, C. J., and Smith, M. M. (2002) Two tarantula peptides inhibit activation of multiple sodium channels. Biochemistry 41, 14734-14747 125. Larabell, C. A., and Chandler, D. E. (1989) The coelomic envelope of Xenopus laevis eggs: a quick-freeze, deep-etch analysis. Developmental biology 131, 126-135 126. Solletti, J. M., Kasas, S., Bertrand, D., and Weisenhorn, A. L. (1994) Atomic force and scanning electron microscopy of Xenopus laevis oocytes. JVSTB 12, 1535-1538 127. Terhag, J., Cavara, N. A., and Hollmann, M. (2010) Cave Canalem: how endogenous ion channels may interfere with heterologous expression in Xenopus oocytes. Methods 51, 66-74

143 128. Aoshima, H., Iio, H., and Kobayashi, S. (1986) Li+ uptake into Xenopus and Cynops oocytes injected with exogenous mRNA, observed by flame emission spectroscopy. Analytical biochemistry 156, 257-262 129. Kobayashi, S., Iio, H., and Aoshima, H. (1986) New translation system of mRNA coding for neurotransmitter receptors using oocytes of the newt, Cynops pyrrhogaster. Brain research 387, 93-96 130. Vargas, R. A., Botero, L., Lagos, L., and Camacho, M. (2004) Bufo marinus oocytes as a model for ion channel protein expression and functional characterization for electrophysiological studies. Cellular physiology and biochemistry : international journal of experimental cellular physiology, biochemistry, and pharmacology 14, 197- 202 131. Markovich, D., and Regeer, R. R. (1999) Expression of membrane transporters in cane toad Bufo marinus oocytes. J Exp Biol 202, 2217-2223 132. Schroeder, B. C., Cheng, T., Jan, Y. N., and Jan, L. Y. (2008) Expression cloning of TMEM16A as a calcium-activated chloride channel subunit. Cell 134, 1019-1029 133. Twiner, M. J., Doucette, G. J., Rasky, A., Huang, X. P., Roth, B. L., and Sanguinetti, M. C. (2012) Marine algal toxin azaspiracid is an open-state blocker of hERG potassium channels. Chemical research in toxicology 25, 1975-1984 134. Nguyen, H. M., and Goldin, A. L. (2010) Sodium channel carboxyl-terminal residue regulates fast inactivation. J Biol Chem 285, 9077-9089 135. Jones, D. K., Peters, C. H., Tolhurst, S. A., Claydon, T. W., and Ruben, P. C. (2011) Extracellular proton modulation of the cardiac voltage-gated sodium channel, Nav1.5. Biophys J 101, 2147-2156 136. Zhang, M. M., Wilson, M. J., Azam, L., Gajewiak, J., Rivier, J. E., Bulaj, G., Olivera, B. M., and Yoshikami, D. (2013) Co-expression of Na(V)beta subunits alters the kinetics of inhibition of voltage-gated sodium channels by pore-blocking mu- conotoxins. Br J Pharmacol 168, 1597-1610 137. Arun, T., Tomassini, V., Sbardella, E., de Ruiter, M. B., Matthews, L., Leite, M. I., Gelineau-Morel, R., Cavey, A., Vergo, S., Craner, M., Fugger, L., Rovira, A., Jenkinson, M., and Palace, J. (2013) Targeting ASIC1 in primary progressive multiple sclerosis: evidence of neuroprotection with amiloride. Brain 136, 106-115 138. Koehn, L. M., Noor, N. M., Dong, Q., Er, S. Y., Rash, L. D., King, G. F., Dziegielewska, K. M., Saunders, N. R., and Habgood, M. D. (2016) Selective inhibition of ASIC1a

144 confers functional and morphological neuroprotection following traumatic spinal cord injury. F1000Res 5, 1822 139. Cristofori-Armstrong, B., and Rash, L. D. (2017) Acid-sensing ion channel (ASIC) structure and function: Insights from spider, snake and sea anemone venoms. Neuropharmacology 127, 173-184 140. Liu, Y., Hagan, R., and Schoellerman, J. (2018) Dual actions of Psalmotoxin at ASIC1a and ASIC2a heteromeric channels (ASIC1a/2a). Sci Rep 8, 7179 141. MacLean, D. M., and Jayaraman, V. (2017) Deactivation kinetics of acid-sensing ion channel 1a are strongly pH-sensitive. Proc Natl Acad Sci U S A 114, E2504-E2513 142. Waxman, S. G., and Zamponi, G. W. (2014) Regulating excitability of peripheral afferents: emerging ion channel targets. Nat Neurosci 17, 153-163 143. da Silva Serra, I., Husson, Z., Bartlett, J. D., and Smith, E. S. (2016) Characterization of cutaneous and articular sensory neurons. Mol Pain 12 144. Steen, K. H., Reeh, P. W., Anton, F., and Handwerker, H. O. (1992) Protons selectively induce lasting excitation and sensitization to mechanical stimulation of nociceptors in rat skin, in vitro. J Neurosci 12, 86-95 145. Yu, Y., Chen, Z., Li, W. G., Cao, H., Feng, E. G., Yu, F., Liu, H., Jiang, H., and Xu, T. L. (2010) A nonproton ligand sensor in the acid-sensing ion channel. Neuron 68, 61- 72 146. Lee, J. Y. P., Saez, N. J., Cristofori-Armstrong, B., Anangi, R., King, G. F., Smith, M. T., and Rash, L. D. (2018) Inhibition of acid-sensing ion channels by diminazene and APETx2 evoke partial and highly variable antihyperalgesia in a rat model of inflammatory pain. Br J Pharmacol 175, 2204-2218 147. Yoder, N., and Gouaux, E. (2018) Divalent cation and chloride ion sites of chicken acid sensing ion channel 1a elucidated by x-ray crystallography. PLoS One 13, e0202134 148. Zuo, Z., Smith, R. N., Chen, Z., Agharkar, A. S., Snell, H. D., Huang, R., Liu, J., and Gonzales, E. B. (2018) Identification of a unique Ca(2+)-binding site in rat acid- sensing ion channel 3. Nat Commun 9, 2082 149. van Zundert, G. C. P., Rodrigues, J., Trellet, M., Schmitz, C., Kastritis, P. L., Karaca, E., Melquiond, A. S. J., van Dijk, M., de Vries, S. J., and Bonvin, A. (2016) The HADDOCK2.2 Web Server: User-Friendly Integrative Modeling of Biomolecular Complexes. J Mol Biol 428, 720-725

145 150. de Vries, S. J., van Dijk, M., and Bonvin, A. M. (2010) The HADDOCK web server for data-driven biomolecular docking. Nat Protoc 5, 883-897 151. Chu, X. P., Wemmie, J. A., Wang, W. Z., Zhu, X. M., Saugstad, J. A., Price, M. P., Simon, R. P., and Xiong, Z. G. (2004) Subunit-dependent high-affinity zinc inhibition of acid-sensing ion channels. J Neurosci 24, 8678-8689 152. Baron, A., Schaefer, L., Lingueglia, E., Champigny, G., and Lazdunski, M. (2001) Zn2+ and H+ are coactivators of acid-sensing ion channels. J. Biol. Chem. 276, 35361- 35367 153. Jiang, Q., Inoue, K., Wu, X., Papasian, C. J., Wang, J. Q., Xiong, Z. G., and Chu, X. P. (2011) Cysteine 149 in the extracellular finger domain of acid-sensing ion channel 1b subunit is critical for zinc-mediated inhibition. Neuroscience 193, 89-99 154. Vukicevic, M., Weder, G., Boillat, A., Boesch, A., and Kellenberger, S. (2006) Trypsin cleaves acid-sensing ion channel 1a in a domain that is critical for channel gating. J Biol Chem 281, 714-722 155. Clark, E. B., Jovov, B., Rooj, A. K., Fuller, C. M., and Benos, D. J. (2010) Proteolytic cleavage of human acid-sensing ion channel 1 by the serine protease matriptase. J Biol Chem 285, 27130-27143 156. Verkest, C., Piquet, E., Diochot, S., Dauvois, M., Lanteri-Minet, M., Lingueglia, E., and Baron, A. (2018) Effects of systemic inhibitors of acid-sensing ion channels 1 (ASIC1) against acute and chronic mechanical allodynia in a rodent model of migraine. Br J Pharmacol 175, 4154-4166 157. Wen, M., Guo, X., Sun, P., Xiao, L., Li, J., Xiong, Y., Bao, J., Xue, T., Zhang, L., and Tian, C. (2015) Site-specific fluorescence spectrum detection and characterization of hASIC1a channels upon toxin mambalgin-1 binding in live mammalian cells. Chem. Commun. (Camb) 51, 8153-8156 158. Besson, T., Lingueglia, E., and Salinas, M. (2017) Pharmacological modulation of Acid-Sensing Ion Channels 1 a and 3 by amiloride and 2-guanidine-4- methylquinazoline (GMQ). Neuropharmacology 125, 429-440 159. Aissouni, Y., El Guerrab, A., Hamieh, A., Ferrier, J., Chalus, M., Lemaire, D., Gregoire, S., Etienne, M., Eschalier, A., Ardid, D., Lingueglia, E., and Marchand, F. (2017) Acid-Sensing Ion Channel 1a in the amygdala is involved in pain and anxiety- related behaviours associated with arthritis. Sci Rep-Uk 7

146 160. Lan, H., Wu, K., Zheng, Y., Pan, M., Huang, Y. C., Gao, S., Zheng, Q. Y., Zheng, J. S., Li, Y. M., Xiao, B. L., and Liu, L. (2016) Total synthesis of mambalgin-1/2/3 by two-segment hydrazide-based native chemical ligation. J Pept Sci 22, 320-326 161. Chen, C. C., England, S., Akopian, A. N., and Wood, J. N. (1998) A sensory neuron- specific, proton-gated ion channel. P Natl Acad Sci USA 95, 10240-10245 162. Ugawa, S., Ueda, T., Ishida, Y., Nishigaki, M., Shibata, Y., and Shimada, S. (2002) Amiloride-blockable acid-sensing ion channels are leading acid sensors expressed in human nociceptors. J Clin Invest 110, 1185-1190 163. Jones, N. G., Slater, R., Cadiou, H., McNaughton, P., and McMahon, S. B. (2004) Acid-induced pain and its modulation in humans. J Neurosci 24, 10974-10979 164. Bagal, S. K., Brown, A. D., Cox, P. J., Omoto, K., Owen, R. M., Pryde, D. C., Sidders, B., Skerratt, S. E., Stevens, E. B., Storer, R. I., and Swain, N. A. (2013) Ion channels as therapeutic targets: a drug discovery perspective. J Med Chem 56, 593-624 165. Klint, J. K., Senff, S., Rupasinghe, D. B., Er, S. Y., Herzig, V., Nicholson, G. M., and King, G. F. (2012) Spider-venom peptides that target voltage-gated sodium channels: pharmacological tools and potential therapeutic leads. Toxicon 60, 478-491 166. Saez, N. J., Senff, S., Jensen, J. E., Er, S. Y., Herzig, V., Rash, L. D., and King, G. F. (2010) Spider-venom peptides as therapeutics. Toxins (Basel) 2, 2851-2871 167. Nebe, J., Ebersberger, A., Vanegas, H., and Schaible, H. G. (1999) Effects of omega- agatoxin IVA, a P-type calcium channel antagonist, on the development of spinal neuronal hyperexcitability caused by knee inflammation in rats. J Neurophysiol 81, 2620-2626 168. Osteen, J. D., Herzig, V., Gilchrist, J., Emrick, J. J., Zhang, C., Wang, X., Castro, J., Garcia-Caraballo, S., Grundy, L., Rychkov, G. Y., Weyer, A. D., Dekan, Z., Undheim, E. A., Alewood, P., Stucky, C. L., Brierley, S. M., Basbaum, A. I., Bosmans, F., King, G. F., and Julius, D. (2016) Selective spider toxins reveal a role for the Nav1.1 channel in mechanical pain. Nature 534, 494-499 169. Siemens, J., Zhou, S., Piskorowski, R., Nikai, T., Lumpkin, E. A., Basbaum, A. I., King, D., and Julius, D. (2006) Spider toxins activate the capsaicin receptor to produce inflammatory pain. Nature 444, 208-212 170. Boscardin, E., Alijevic, O., Hummler, E., Frateschi, S., and Kellenberger, S. (2016) The function and regulation of acid-sensing ion channels (ASICs) and the epithelial Na(+) channel (ENaC): IUPHAR Review 19. Br J Pharmacol 173, 2671-2701

147 171. Yermolaieva, O., Leonard, A. S., Schnizler, M. K., Abboud, F. M., and Welsh, M. J. (2004) Extracellular acidosis increases neuronal cell calcium by activating acid- sensing ion channel 1a. Proc Natl Acad Sci U S A 101, 6752-6757 172. Duan, B., Wu, L. J., Yu, Y. Q., Ding, Y., Jing, L., Xu, L., Chen, J., and Xu, T. L. (2007) Upregulation of acid-sensing ion channel ASIC1a in spinal dorsal horn neurons contributes to inflammatory pain hypersensitivity. J Neurosci 27, 11139-11148 173. Vergo, S., Craner, M. J., Etzensperger, R., Attfield, K., Friese, M. A., Newcombe, J., Esiri, M., and Fugger, L. (2011) Acid-sensing ion channel 1 is involved in both axonal injury and demyelination in multiple sclerosis and its animal model. Brain 134, 571- 584 174. Ziemann, A. E., Schnizler, M. K., Albert, G. W., Severson, M. A., Howard, M. A., 3rd, Welsh, M. J., and Wemmie, J. A. (2008) Seizure termination by acidosis depends on ASIC1a. Nat. Neurosci. 11, 816–822 175. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W., and Lipman, D. J. (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 25, 3389-3402 176. Kuduk, S. D., Chang, R. K., Wai, J. M., Di Marco, C. N., Cofre, V., DiPardo, R. M., Cook, S. P., Cato, M. J., Jovanovska, A., Urban, M. O., Leitl, M., Spencer, R. H., Kane, S. A., Hartman, G. D., and Bilodeau, M. T. (2009) Amidine derived inhibitors of acid-sensing ion channel-3 (ASIC3). Bioorg Med Chem Lett 19, 4059-4063 177. Kuduk, S. D., Di Marco, C. N., Bodmer-Narkevitch, V., Cook, S. P., Cato, M. J., Jovanovska, A., Urban, M. O., Leitl, M., Sain, N., Liang, A., Spencer, R. H., Kane, S. A., Hartman, G. D., and Bilodeau, M. T. (2010) Synthesis, structure-activity relationship, and pharmacological profile of analogs of the ASIC-3 inhibitor A-317567. ACS chemical neuroscience 1, 19-24 178. King, G. F., Gentz, M. C., Escoubas, P., and Nicholson, G. M. (2008) A rational nomenclature for naming peptide toxins from spiders and other venomous animals. Toxicon 52, 264-276 179. Sherwood, T., Franke, R., Conneely, S., Joyner, J., Arumugan, P., and Askwith, C. (2009) Identification of protein domains that control proton and calcium sensitivity of ASIC1a. J Biol Chem 284, 27899-27907 180. Herzig, V., and King, G. F. (2015) The Cystine Knot Is Responsible for the Exceptional Stability of the Insecticidal Spider Toxin omega-Hexatoxin-Hv1a. Toxins (Basel) 7, 4366-4380

148 181. Kikuchi, K., Sugiura, M., and Kimura, T. (2015) High Proteolytic Resistance of Spider- Derived Inhibitor Cystine Knots. Int J Pept 2015, 537508 182. Olsen, J. V., Ong, S. E., and Mann, M. (2004) Trypsin cleaves exclusively C-terminal to arginine and lysine residues. Mol Cell Proteomics 3, 608-614 183. Kaas, Q., Westermann, J. C., Halai, R., Wang, C. K., and Craik, D. J. (2008) ConoServer, a database for conopeptide sequences and structures. Bioinformatics 24, 445-446 184. Lebbe, E. K., Ghequire, M. G., Peigneur, S., Mille, B. G., Devi, P., Ravichandran, S., Waelkens, E., D'Souza, L., De Mot, R., and Tytgat, J. (2016) Novel Conopeptides of Largely Unexplored Indo Pacific Conus sp. Mar Drugs 14 185. Kaas, Q., Yu, R., Jin, A. H., Dutertre, S., and Craik, D. J. (2012) ConoServer: updated content, knowledge, and discovery tools in the conopeptide database. Nucleic Acids Res 40, D325-330 186. Imperial, J. S., Chen, P., Sporning, A., Terlau, H., Daly, N. L., Craik, D. J., Alewood, P. F., and Olivera, B. M. (2008) Tyrosine-rich conopeptides affect voltage-gated K+ channels. J Biol Chem 283, 23026-23032 187. Huang, C. C., Lyu, P. C., Lin, C. H., and Hsu, K. S. (1997) Conantokin-T selectively antagonizes N-methyl-D-aspartate-evoked responses in rat hippocampal slice. Toxicon 35, 355-363 188. Platt, R. J., Curtice, K. J., Twede, V. D., Watkins, M., Gruszczynski, P., Bulaj, G., Horvath, M. P., and Olivera, B. M. (2014) From molecular phylogeny towards differentiating pharmacology for NMDA receptor subtypes. Toxicon 81, 67-79 189. Reyes-Guzman, E. A., Vega-Castro, N., Reyes-Montano, E. A., and Recio-Pinto, E. (2017) Antagonistic action on NMDA/GluN2B mediated currents of two peptides that were conantokin-G structure-based designed. BMC Neurosci 18, 44 190. Twede, V. D., Teichert, R. W., Walker, C. S., Gruszczynski, P., Kazmierkiewicz, R., Bulaj, G., and Olivera, B. M. (2009) Conantokin-Br from Conus brettinghami and selectivity determinants for the NR2D subunit of the NMDA receptor. Biochemistry 48, 4063-4073 191. Wittekindt, B., Malany, S., Schemm, R., Otvos, L., Maccecchini, M. L., Laube, B., and Betz, H. (2001) Point mutations identify the glutamate binding pocket of the N-methyl- D-aspartate receptor as major site of conantokin-G inhibition. Neuropharmacology 41, 753-761

149 192. Biggs, J. S., Rosenfeld, Y., Shai, Y., and Olivera, B. M. (2007) Conolysin-Mt: a conus peptide that disrupts cellular membranes. Biochemistry 46, 12586-12593 193. Craig, A. G., Norberg, T., Griffin, D., Hoeger, C., Akhtar, M., Schmidt, K., Low, W., Dykert, J., Richelson, E., Navarro, V., Mazella, J., Watkins, M., Hillyard, D., Imperial, J., Cruz, L. J., and Olivera, B. M. (1999) Contulakin-G, an O-glycosylated invertebrate neurotensin. J Biol Chem 274, 13752-13759 194. Maillo, M., Aguilar, M. B., Lopez-Vera, E., Craig, A. G., Bulaj, G., Olivera, B. M., and Heimer de la Cotera, E. P. (2002) Conorfamide, a Conus venom peptide belonging to the RFamide family of neuropeptides. Toxicon 40, 401-407 195. Wu, X., Shao, X., Guo, Z. Y., and Chi, C. W. (2010) Identification of neuropeptide Y- like conopeptides from the venom of Conus betulinus. Acta Biochim Biophys Sin (Shanghai) 42, 502-505 196. Robinson, S. D., Safavi-Hemami, H., Raghuraman, S., Imperial, J. S., Papenfuss, A. T., Teichert, R. W., Purcell, A. W., Olivera, B. M., and Norton, R. S. (2015) Discovery by proteogenomics and characterization of an RF-amide neuropeptide from cone snail venom. J Proteomics 114, 38-47 197. Campos-Lira, E., Carrillo, E., Aguilar, M. B., Gajewiak, J., Gomez-Lagunas, F., and Lopez-Vera, E. (2017) Conorfamide-Sr3, a structurally novel specific inhibitor of the Shaker K(+) channel. Toxicon 138, 53-58 198. Wonnacott, S., Bermudez, I., Millar, N. S., and Tzartos, S. J. (2018) Nicotinic acetylcholine receptors. Br J Pharmacol 175, 1785-1788 199. Giribaldi, J., and Dutertre, S. (2018) alpha-Conotoxins to explore the molecular, physiological and pathophysiological functions of neuronal nicotinic acetylcholine receptors. Neurosci Lett 679, 24-34 200. Prashanth, J. R., Dutertre, S., Jin, A. H., Lavergne, V., Hamilton, B., Cardoso, F. C., Griffin, J., Venter, D. J., Alewood, P. F., and Lewis, R. J. (2016) The role of defensive ecological interactions in the evolution of conotoxins. Mol Ecol 25, 598-615 201. Abraham, N., and Lewis, R. J. (2018) Neuronal Nicotinic Acetylcholine Receptor Modulators from Cone Snails. Mar Drugs 16 202. Vetter, I., Mozar, C. A., Durek, T., Wingerd, J. S., Alewood, P. F., Christie, M. J., and Lewis, R. J. (2012) Characterisation of Na(v) types endogenously expressed in human SH-SY5Y neuroblastoma cells. Biochem Pharmacol 83, 1562-1571 203. Vetter, I. (2012) Development and optimization of FLIPR high throughput calcium assays for ion channels and GPCRs. Adv Exp Med Biol 740, 45-82

150 204. Pennington, M. W., Czerwinski, A., and Norton, R. S. (2018) Peptide therapeutics from venom: Current status and potential. Bioorg Med Chem 26, 2738-2758 205. Durek, T., and Craik, D. J. (2015) Therapeutic conotoxins: a US patent literature survey. Expert Opin Ther Pat 25, 1159-1173 206. Vetter, I., and Lewis, R. J. (2012) Therapeutic potential of cone snail venom peptides (conopeptides). Curr Top Med Chem 12, 1546-1552 207. Starobova, H., S, W. A. H., Lewis, R. J., and Vetter, I. (2018) Transcriptomics in pain research: insights from new and old technologies. Mol Omics 208. Vetter, I., Davis, J. L., Rash, L. D., Anangi, R., Mobli, M., Alewood, P. F., Lewis, R. J., and King, G. F. (2011) Venomics: a new paradigm for natural products-based drug discovery. Amino acids 40, 15-28 209. Elphick, M. R., and Mirabeau, O. (2014) The Evolution and Variety of RFamide-Type Neuropeptides: Insights from Deuterostomian Invertebrates. Front Endocrinol (Lausanne) 5, 93 210. Price, D. A., and Greenberg, M. J. (1977) Structure of a molluscan cardioexcitatory neuropeptide. Science 197, 670-671 211. Lingueglia, E., Champigny, G., Lazdunski, M., and Barbry, P. (1995) Cloning of the amiloride-sensitive FMRFamide peptide-gated sodium channel. Nature 378, 730-733 212. Askwith, C. C., Cheng, C., Ikuma, M., Benson, C., Price, M. P., and Welsh, M. J. (2000) Neuropeptide FF and FMRFamide potentiate acid-evoked currents from sensory neurons and proton-gated DEG/ENaC channels. Neuron 26, 133-141 213. Bargeton, B., Iwaszkiewicz, J., Bonifacio, G., Roy, S., Zoete, V., and Kellenberger, S. (2019) Mutations in the palm domain disrupt modulation of acid-sensing ion channel 1a currents by neuropeptides. Sci Rep 9, 2599 214. Cottrell, G. A., Jeziorski, M. C., and Green, K. A. (2001) Location of a ligand recognition site of FMRFamide-gated Na(+) channels. FEBS Lett 489, 71-74 215. Niu, Y. Y., Yang, Y., Liu, Y., Huang, L. D., Yang, X. N., Fan, Y. Z., Cheng, X. Y., Cao, P., Hu, Y. M., Li, L., Lu, X. Y., Tian, Y., and Yu, Y. (2016) Exploration of the Peptide Recognition of an Amiloride-sensitive FMRFamide Peptide-gated Sodium Channel. J Biol Chem 291, 7571-7582 216. Yang, X. N., Niu, Y. Y., Liu, Y., Yang, Y., Wang, J., Cheng, X. Y., Liang, H., Wang, H. S., Hu, Y. M., Lu, X. Y., Zhu, M. X., Xu, T. L., Tian, Y., and Yu, Y. (2017) The nonproton ligand of acid-sensing ion channel 3 activates mollusk-specific FaNaC

151 channels via a mechanism independent of the native FMRFamide peptide. J Biol Chem 292, 21662-21675

152 APPENDIX I

UQ Research and Innovation Director, Research Management Office Nicole Thompson

ANIMAL ETHICS APPROVAL CERTIFICATE 17-Dec-2013

Activity Details Chief Investigator: Professor Joseph Lynch, Queensland Brain Institute Title: Discovering new drugs and understanding the mechanisms by which drugs work at neuronal ion channels AEC Approval Number: QBI/059/13/ARC/NHMRC Previous AEC Number: QBI/017/10/ARC (NF) Approval Duration: 26-Apr-2013 to 26-Apr-2016 Funding Body: ARC, NHMRC Group: Anatomical Biosciences Other Staff/Students: Glenn King, Ben Cristofori-Armstrong, Ming Soh, Sharifun Islam, Lachlan Rash, Han Lu, Anna Bode, Irene Chassagnon, Julie Klint, Darshani Rupasinghe, Robiul Islam, Fernanda Caldas Cardoso, Sahil Talwar, Angelo Keramidas

Location(s): St Lucia Bldg 75 - AIBN St Lucia Bldg 81 - Otto Hirschfeld

Summary Subspecies Strain Class Gender Source Approved Remaining Amphibians Xenopus laevis Adults Mix Commercial 125 125 breeding colony

Permit(s):

Proviso(s): (1) a) Xenopus frogs must have permanent identification marks and be housed according to the DNR permit regulations; b) The CI must ensure that the frogs are well identified and accurate records are kept of their use. (2) The frogs may be used for collection of oocytes NO more than 6 times.

Approval Details

Description Amount Balance

Amphibians (Xenopus laevis, Mix, Adults, Commercial breeding colony) 10 Apr 2013 Initial approval 65 65 11 Dec 2013 Modification #2 60 125

UQ Research and Innovation Cumbrae-Stewart Building T +61 7 3365 2925 (Enquiries) E [email protected] The University of Queensland Research Road T +61 7 3365 2713 (Manager) W www.uq.edu.au/research/rid/ Brisbane Qld 4072 Australia F +61 7 3365 4455

Page 1 of 2

153

Please note the animal numbers supplied on this certificate are the total allocated for the approval duration

Please use this Approval Number: 1. When ordering animals from Animal Breeding Houses 2. For labelling of all animal cages or holding areas. In addition please include on the label, Chief Investigator's name and contact phone number. 3. When you need to communicate with this office about the project.

It is a condition of this approval that all project animal details be made available to Animal House OIC. (UAEC Ruling 14/12/2001)

The Chief Investigator takes responsibility for ensuring all legislative, regulatory and compliance objectives are satified for this project. This certificate supercedes all preceeding certificates for this project (i.e. those certificates dated before 17-Dec-2013)

UQ Research and Innovation Cumbrae-Stewart Building T +61 7 3365 2925 (Enquiries) E [email protected] The University of Queensland Research Road T +61 7 3365 2713 (Manager) W www.uq.edu.au/research/rid/ Brisbane Qld 4072 Australia F +61 7 3365 4455

Page 2 of 2

154

UQ Research and Innovation Director, Research Management Office Nicole Thompson Animal Ethics Approval Certificate 20-Apr-2016 Please check all details below and inform the Animal Welfare Unit within 10 working days if anything is incorrect.

Activity Details Chief Investigator: Professor Joseph Lynch, Queensland Brain Institute Title: Investigating the effects of new drugs for pain, stroke and anxiety on neuronal ion channels AEC Approval Number: QBI/AIBN/087/16/NHMRC/ARC Previous AEC Number: Approval Duration: 20-Apr-2016 to 20-Apr-2019 Funding Body: ARC, NHMRC Group: Anatomical Biosciences Other Staff/Students: Lachlan Rash, Natalie Saez, Irene Chassagnon, Angelo Keramidas, Mohammed Atif, Jessie Er, Linlin Ma, Ming Soh, Ben Cristofori-Armstrong, Argel Estrada, Nela Durisic, Glenn King, Jennifer Smith, Sharifun Islam Location(s): St Lucia Bldg 75 - AIBN

Summary Subspecies Strain Class Gender Source Approved Remaining Amphibians Xenopus Laevis Adults Female 58 58 (breeder) Amphibians Marsabit Clawed Adults Female 22 22 Frog (Xenopus borealis) (breed)

Permits

Provisos Pest Species (Xenopus frogs) Proviso: a) Xenopus frogs must have permanent identification marks and be housed according to the DNR permit regulations b) The CI must ensure that the toads are well identified and accurate records are kept of their use.

The CI is required to ensure that no animal is used for oocyte collection more than 6 times.

Approval Details

Description Amount Balance

Amphibians (Marsabit Clawed Frog (Xenopus borealis) (breed), Female, Adults, ) 20 Apr 2016 Initial approval 22 22 Amphibians (Xenopus Laevis (breeder), Female, Adults, ) 20 Apr 2016 Initial approval 58 58

Animal Welfare Unit Cumbrae-Stewart Building +61 7 336 52925 (Enquiries) [email protected] UQ Research and Innovation Research Road +61 7 334 68710 (Enquiries) uq.edu.au/research The University of Queensland Brisbane Qld 4072 Australia +61 7 336 52713 (Coordinator) Page 1 of 2

155

156